EP1366067A2

EP1366067A2 - Expression technology for proteins containing a hybrid isotype antibody moiety

Info

Publication number: EP1366067A2
Application number: EP02717581A
Authority: EP
Inventors: Stephen D. Gillies; Jeffrey Way; Kin-Ming Lo
Original assignee: Merck Patent GmbH
Current assignee: Merck Patent GmbH
Priority date: 2001-03-07
Filing date: 2002-03-07
Publication date: 2003-12-03
Anticipated expiration: 2022-03-07
Also published as: WO2002072605A3; AU2002248571B2; KR100900176B1; CA2440221C; EP1366067B1; DK1366067T3; ZA200307792B; WO2002072605A2; PL366981A1; CN1330664C; ES2393733T3; EP1366067A4; MXPA03008031A; BR0207854A; US7148321B2; CN1509293A; JP2004525630A; JP4234438B2; KR20090010127A; CA2440221A1

Abstract

Disclosed are methods and compositions for efficiently expressing antibody fusion proteins. Antibody fusion proteins of the invention include a hybrid antibody moiety containing sequences from more than one type of antibody and/or mutant antibody sequences. Hybrid antibody fusion proteins of the invention may be produced at high levels and may combine functional properties characteristic of different antibody types in addition to functional properties of a non-antibody moiety.

Description

EXPRESSION TECHNOLOGY FOR PROTEINS CONTAINING A HYBRID

ISOTYPE ANTIBODY MOIETY

Related Applications [0001 ] This application claims priority to, and the benefit of U.S.S.N. 60/274,096 filed March 7, 2001 , the disclosure of which is incoφorated by reference herein in its entirety.

Field of the Invention

[0002] The present invention relates generally to methods and compositions for antibodies with moieties deriving from two or more isotypes, and fusion proteins derived therefrom, including proteins containing an antibody moiety with altered effector functions, increased protein expression and/or decreased oligomerization. The invention particularly relates to antibodies and fusion proteins in which a hinge region derives from one isotype and the CH2 domain derives from a different isotype.

Background of the Invention [0003] The efficiency of protein expression from genetically engineered cells is an important commercial concern. Some commercially important proteins are best made from eukaryotic cells, such as mammalian cells, plant cells, or yeast cells, in order to ensure correct folding and glycosylation. However, the cost of maintaining large cultures of eukaryotic cells means that proteins produced in this way are expensive to make. Therefore, there is a need in the art to maximize the expression levels of proteins from eukaryotic cells.

[0004] A related issue is that therapeutic proteins produced from eukaryotic cells must be expressed in the correct conformational state. Normally, mechanisms of transcription and translation ensure that a genetically engineered cell will produce a protein whose sequence is determined by the nucleic acid encoding the protein.

However, after transcription and translation, the protein may fail to fold properly and may be degraded. Alternatively, a protein may be produced in an aggregated state, such that activity is reduced. Even if an aggregated protein is active, it may be pharmacologically unacceptable due to increased immunogenicity compared to a non- aggregated protein. Thus, a pharmacologically acceptable protein preparation should generally be substantially free of aggregated proteins. [0005] The amount of a protein that is expressed from a genetically engineered eukaryotic cell is a function of the rate of transcription of the encoding gene, the efficiency of mRNA splicing and of export from the nucleus, and the efficiency of translation. The role that these events play in protein expression is sufficiently well understood that one skilled in the art of genetic engineering and protein expression can generally incoφorate appropriate nucleic acid sequences into the design of an expression construct with efficient transcription, splicing, mRNA export, and translation.

[0006] However, the amount of a correctly folded, non-aggregated protein that is produced from a eukaryotic cell is also a function of the amino acid sequence of the protein, as well as the nucleic acid sequences that determine transcription, splicing, mRNA export, translation, and post-translational modification. For example, it is thought that a significant fraction of proteins synthesized in a cell is degraded. The features in a protein that determine whether or not it should be degraded are currently subject to intensive study, but presently it is not possible to predict the efficiency of protein folding, degradation, or aggregation by simply examining the sequence of a protein. Some naturally occurring proteins fold efficiently, are resistant to proteolysis, and do not aggregate. In contrast, other proteins fold inefficiently, are rapidly degraded, and aggregate. [0007] Antibodies and artificial proteins containing a portion of an antibody, termed antibody fusion proteins or lg fusion proteins herein, are useful for a variety of puφoses relating to the targeting capability of antibody variable domains as well as the ability of the constant regions to bind to various other proteins. Antibody and antibody fusion protein preparations are particularly useful when they are correctly folded and non-aggregated. Therefore there is a need in the art for methods and compositions for the production of antibody and antibody fusion protein preparations with reduced aggregation.

[0008] Additionally, antibodies and antibody fusion proteins are useful since their ability to bind to various other proteins enables them, for example, to elicit specific effector functions. In some instances specific effector functions are desirable but often the loss of effector functions is preferable. The antibody component of a fusion protein may be altered to reduce or eliminate effector functions by utilizing a modified antibody. Antibody and antibody fusion protein preparations are also useful when they are modified to alter functionality. Therefore there is a need in the art for methods and compositions for the production of modified antibodies and antibody fusion proteins with altered effector functions.

[0009] Protein drugs can be degraded by proteases, such that their delivery and pharmacokinetic properties are suboptimal. There is a need in the art for improvement of protein drugs that have the useful properties of certain proteins, but that have greater protease resistance.

Summary of the Invention [0010] The present invention features methods and compositions useful for producing intact antibodies, immunocytokines, immunofusins, immunoligands, and other antibody and Fc fusion proteins that enhance the expression, proper oligomerization, purification, and protease resistance of a desired fusion protein optionally with modified, combined, or decreased Fc effector functions. Specifically, the invention provides antibody moieties with hybrid isotypes, optionally using mutant lg components for use in intact antibodies and in fusion proteins containing an antibody moiety.

IgG/IgG hybrid isotypes

[0011] In one set of preferred embodiments, the invention provides fusion proteins with decreased effector functions and improved assembly. Such fusion proteins are particularly useful when the lg moiety serves to enhance expression and improve serum half-life, but when the immunological functions of the lg moiety are not needed.

[0012] In these embodiments, the fusion protein preferably comprises CHI , CH2, and/or CH3 domains of an IgG2 or IgG4, combined with a hinge region from IgGl or a hinge region from IgG4, the latter hinge region preferentially comprising a mutation that enhances correct formation of disulfide bonds between heavy chain- derived moieties (Angal S, et al. Mol Immunol 1993 Jan;30(l):105-8). Fusion proteins of this embodiment facilitate high-level expression and improve the correct assembly of intact antibodies and lg fusion proteins containing Fc regions.

[0013] In a more preferred embodiment, the fusion proteins also contain one or more mutations in the lg moiety. For example, the lg moiety is mutated to further reduce any remaining effector function that is not desired. For example, the Clq binding site in the CH2 domain of IgG2 is mutant. For example, normal IgG4 does not bind complement because it contains a serine instead of a corresponding proline at position 331 in IgGl (Eu nomenclature)(Tao, MA et al (1993) J. Exp. Med. 178:661- 667; Brekke, OH et al (1994) Eur. J. Immunol. 24:2542-2547); a similar mutation in IgG2 reduces Clq binding. Other residues known to be involved in Clq binding could be modified such as residues at positions 318, 320, 322 and 297 (Duncan AR (1988) Nature: 332:738-740), resulting in a decrease in Clq binding.

[0014] In another set of preferred embodiments, a mutation in the hinge region is also present. For example, in cases where an antibody light chain is also present, a form of an IgGl hinge region with the cysteine residues in the normal number at their normal positions is preferred. However, in cases where an antibody light chain is not present as a distinct polypeptide chain, an IgGl hinge in which the first cysteine is mutated to another residue is preferred. For example, it is useful to employ such a mutated hinge region in Fc-X proteins, X-Fc proteins, and in single-chain antibodies in which the light chain variable region is attached to the heavy chain by means of a polypeptide linker. The first cysteine in the IgGl hinge is preferably mutated to a serine in this context.

[0015] In a second class of embodiments involving a mutant hinge region, a mutant form of the IgG4 hinge that allows efficient disulfide bonding between the two heavy chains is used.

[0016] In a third class of embodiments involving a mutant hinge region, a mutant form of the IgG2 hinge, in which the first two cysteines are each mutated to another amino acid, is used in a hybrid isotype antibody or lg fusion protein. It is also convenient to use such a mutant hinge in an antibody or lg fusion protein that is entirely derived from IgG2. For example, a modified IgG2 hinge with the sequence ERKSSVECPPCP (SEQ ID NO: 1) is used in the context of an antibody or lg fusion protein. Another useful type of hinge is a hybrid between the IgG2 hinge and the IgG4 hinge, such as the sequence ESKYG-VECPPCP (SEQ ID NO: 2), in which the 5 amino acids before the dash are derived from IgG4 and the remaining amino acids are from IgG2. These embodiments are particularly useful in the context of antibodies and lg fusion proteins expressed and secreted from eukaryotic cells, because these hinge embodiments promote correct assembly of the proteins. These embodiments may be used as an alternative to the IgGl hinge. A key feature of these embodiments of an antibody hinge is that they only have two cysteine residues.

[0017] Yet another class of embodiments involve mutations in the lg moiety of hybrid isotype lg fusion proteins at the junction between the lg and non-Ig moieties. In one embodiment, the alterations in the amino acid sequence of the fusion protein are preferentially at the junction of the lg moiety and the non-Ig moiety and preferably lie within 10 amino acids of the junction point. More preferably, the amino acid changes involve changing the C-terminal lysine of the antibody moiety to a hydrophobic amino acid such as alanine or leucine. [0018] An alternative embodiment is useful in circumstances where it is desirable to shorten the half-life of a fusion protein. IgG3 has a short half-life relative to other IgG isotypes due to a modification in the FcRn/FcRp binding site located in the CH3 domain (H435 to R)(see Ward, ES. and Gheti, V [1995] Therapeutic Immunology 2:77-94). A fusion protein with IgG3 CH2 and CH3 domains may be used when short term exposure is desirable. According to this embodiment, it is useful to use an IgG3 CH3 domain in combination with an IgGl hinge; such an IgGl(hinge)-IgG3(CH3) fusion protein has superior expression and assembly properties compared to an lg fusion protein containing an IgG3 hinge and an IgG3 CH3 domain.

[0019] In a more preferred embodiment of an lg fusion protein designed to have a short serum half-life, reduced effector functions, and efficient assembly, a hybrid lg region is used in which the hinge region derives from IgGl, the CH2 domain derives from IgG2, and the CH3 domain derives from IgG3.

IgG/IgA hybrid isotypes [0020] A distinct embodiment of the invention provides hybrid isotype lg fusion proteins with enhanced protease resistance and enhanced serum half-life. This embodiment is particularly useful in situations in which an lg fusion protein is exposed to an environment rich in proteases, such as the gut or another mucosal tissue, for example during oral delivery of a lg fusion protein drug. According to this embodiment, an lg fusion protein containing elements of the IgG and IgA constant regions is provided. In a preferred embodiment, the hinge of IgAl and the CH2 and CH3 domains of an IgG are used. In a distinct preferred embodiment, the segments of amino acids encoding the O-linked glycosylation sites in the Fc region of IgA are grafted into the Fc region of an IgG.

IgG/IgM hybrid isotypes [0021] Yet another embodiment of the invention provides hybrid isotype antibodies and lg fusion proteins with oligomerization features of IgA or IgM but with effector functions characteristic of IgG. For example, a protein comprising the hinge region and CH2 domain of IgGl or IgG3 fused to the CH3 and CH4 domain of IgM is provided. In a more preferred embodiment, an antibody comprising heavy and light chain variable regions and also comprising an IgG hinge and CH2 region fused to the CH3 and CH4 domain of IgM is provided. In an alternative more preferred embodiment, an lg fusion protein of the form X-Fc is provided, in which X is preferably a ligand for a cell-surface receptor, and the Fc moiety comprises the CH3 and CH4 domain of IgM. Such a molecule combines the ADCC effector function of an IgG with the high valency of IgM.

[0022] In a preferred embodiment of a hybrid isotypes using the CH4 domain of IgM or IgA, the C-terminal cysteine in the CH4 domain is mutated to block disulfide bonding to the J chain. This reduces secretion of the lg fusion protein into the gut.

Preferred non-Ig moieties

[0023] The preferred type of non-Ig moiety in the fusion proteins of the invention is a protein or moiety that is normally extracellular when not part of an lg fusion protein. For example, a hormone, cytokine, chemokine, secreted enzyme, or extracellular portion of a trans-membrane receptor. [0024] In a preferred embodiment, the non-immunoglobulin component is a protein such as an anti-obesity protein. For example, the non-immunoglobulin component is leptin, CNTF, CLC/CLF-1 or a portion of Acφ30.

[0025] In yet another preferred embodiment, the non-immunoglobulin component is a protein such as erythropoeitin or EPO. [0026] In an alternative embodiment, the non-immunoglobulin component of the fusion protein is a hormone. For example, the non-immunoglobulin component may be insulin, growth hormone, or glucagon-like peptide l(GLP-l). [0027] In another embodiment, the non-immunoglobulin component of the fusion protein is a cytokine. The term "cytokine" is used herein to describe naturally occurring or recombinant proteins, analogs thereof, and fragments thereof that elicit a specific response in a cell which has a receptor for that cytokine. Preferably cytokines are proteins that may be produced and secreted by a cell. Preferably, cytokines include interleukins such as interleukin-2 (IL-2), IL-4, IL-5, IL-6, IL-7, IL- 10, IL-12, IL-13, IL-14, IL-15, IL-16 and IL-18, hematopoietic factors such as granulocyte-macrophage colony stimulating factor (GM-CSF), G-CSF and erythropoietin, tumor necrosis factors (TNF) such as TNFα, lymphokines such as lymphotoxin, regulators of metabolic processes such as leptin, interferons such as interferon α, interferon β, and interferon γ, and chemokines. Preferably, the Ig- cytokine fusion protein of the present invention displays cytokine biological activity.

[0028] In an alternative preferred embodiment, the non-immunoglobulin component of the fusion protein is a ligand-binding protein with biological activity. Such ligand-binding proteins may, for example, (1) block receptor-ligand interactions at the cell surface; or (2) neutralize the biological activity of a molecule (e.g., a cytokine) in the fluid phase of the blood, thereby preventing it from reaching its cellular target. Preferred ligand-binding proteins include CD4, CTLA-4, TNF receptors, or interleukin receptors such as the IL-1 and IL-4 receptors. Preferably, the antibody-receptor fusion protein of the present invention displays the biological activity of the ligand-binding protein. One highly preferred embodiment comprises the extracellular TNF-receptor domain fragment used in the protein drug Enbrel, in the format TNFR-hinge-CH2-CH3 or hinge-CH2-CH3-TNFR, in which the CH2 and CH3 domains are derived from IgG2 or IgG4 and each of the two hinge regions in the dimeric Fc has three or fewer cysteines, and even more preferably, two or fewer cysteines.

[0029] Another type of preferred ligand-binding protein has the capacity to bind to small molecules rather than proteins. For example, it is convenient to fuse avidin to a hybrid isotype lg moiety, such as an antibody. The hybrid isotype antibody- avidin fusion is then administered to a mammal such as a mouse or human and becomes concentrated in a target tissue of the body, as determined by the specificity of a V region of the antibody. After the antibody-avidin fusion protein has sufficiently cleared from the body, a conjugate of biotin and a therapeutic molecule is administered. The biotin conjugate becomes concentrated in the target tissue by virtue of binding to avidin, which results in a lessening of side effects that could arise from a concentration of the therapeutic molecule in non-target tissues. This strategy can be used with other ligand/ligand-binding protein pairs. [0030] Another type ofpreferred non-immunoglobulin moiety is an enzyme. For example, an enzyme with a distinctive specificity may be fused to a hybrid isotype lg moiety, such as an antibody. The hybrid isotype antibody-enzyme fusion is then administered to a mammal such as a mouse or human and becomes concentrated in a target tissue of the body, as determined by the specificity of a V region of the antibody. In one preferred treatment method of the invention, after the antibody- enzyme fusion protein has sufficiently cleared from the body, a prodrug that can be cleaved into an active form by the enzyme is administered. The activated drug becomes concentrated in the target tissue, which results in a lessening of side effects that could arise from a concentration of the activated drug molecule in non-target tissues. In a highly preferred form of this embodiment, the activated drug is an anticancer drug, such as a cytotoxic agent. In an alternative highly preferred embodiment, the enzyme itself has a therapeutic activity. For example, an RNAse such as Onconase is coupled to a hybrid isotype antibody fusion protein and targeted to a tumor via an antibody V region.

Nucleic acids

[0031] This invention also provides novel nucleic acid sequences that facilitate the expression and secretion of lg fusion proteins and intact antibodies with hybrid isotypes, as well as methods for the construction of such nucleic acids. [0032] A particularly useful feature of genomic sequences encoding antibody proteins is that the variable region, CHI, hinge, CH2, CH3, and CH4 regions are encoded by separate exons. This feature facilitates the engineering of hybrid isotype lg fusion proteins by 'exon shuffling' (Zuckier et al., Cancer Research [1998] 58:3905-8; Poon et al., J. Biol. Chem. [1995] 270:8571-7; Jefferis R, et al., Mol Immunol. [1990] 27:1237-40; Chappel MS, Proc Natl Acad Sci U S A. [1991] 88:9036-40; Senior BW, et al., Infect Immun. [2000] 68:463-9.). [0033] For Fc fusion proteins, nucleic acid molecules can encode the protein in various configurations. In a preferred set of embodiments, a nucleic acid molecule encodes, serially in a 5' to 3' direction, (i) a signal sequence, an immunoglobulin Fc region and a target protein sequence or (ii) a signal sequence, a target protein, and an immunoglobulin Fc region, or (iii) a signal sequence, a first target protein, an immunoglobulin Fc region, and a second target protein. The resulting nucleic acid molecule thereby encodes an Fc-X, X-Fc, or X-Fc-Y structure where X and Y are a target protein or proteins. For example, X and Y may themselves be fusion proteins. Linkers are optionally encoded between these moieties. [0034] Similarly, according to the invention, nucleic acids encoding whole antibody lg fusion proteins are designed to encode a signal sequence at the N- terminus of each heavy chain moiety and each light chain moiety.

[0035] A nucleic acid of the invention can be incoφorated in functional association into a replicable expression vector which can then be introduced into a eukaryotic host cell competent to produce a fusion protein. The resultant lg fusion proteins are produced efficiently and secreted from the eukaryotic host cells. Secreted lg fusion proteins may be collected from the culture media without lysing the eukaryotic host cell. The protein product can be assayed for activity and/or purified using common reagents as desired, and/or cleaved from the fusion partner, all using conventional techniques. Alternatively, a nucleic acid of the invention may be introduced into a bacterial cell, and the resulting lg fusion protein purified according to standard techniques.

[0036] The invention also provides methods of enhancing the levels of antibodies and lg fusion proteins produced in cells. The method is preferably applied to production in eukaryotic cells, preferably mammalian cells. For example, according to the method, the production of a starting antibody or lg fusion protein is improved by exchanging nucleic acid sequences encoding domains of the lg moiety with corresponding sequences encoding domains from other antibody isotypes, or with mutant sequences, comparing expression levels by assaying the production of the altered protein as described herein, and choosing particular expression constructs that give the highest levels. This process may be used iteratively. It is particularly useful to interchange hinge regions. Treatment

[0037] The invention also provides methods of treatment using the modified antibodies and lg fusion proteins. Accordingly, the invention provides processes that are both efficient and inexpensive as well as proteins that are less immunogenic. [0038] The foregoing and other objects, features and advantages of the present invention will be made more apparent from the detailed description, drawings, and claims that follow.

Brief Description of the Drawings

[0039] Figures 1 A-D are schematic illustrations of IgG hybrid isotypes used to prepare fusion proteins in accordance with certain aspects of the invention; the thick lines represent disulfide bonds connecting cysteine residues; antibody domains are indicated in Figure A; IgGl domains are shown in black, IgG2 domains are shown in white, and variable and light chain domains are shown striped. [0040] Figure 1 A shows an IgGl isotype.

[0041] Figure IB shows an IgG2 isotype.

[0042] Figure IC shows an isotype hybrid with an IgG2 and an IgGl hinge having a mutation of the first cysteine.

[0043] Figure 1 D shows an IgG hybrid γl (CH 1 -H) γ2(CH2-CH3). [0044] Figures 2A-C are schematic illustrations of lg fusion proteins comprising Fc regions that contain hybrid isotypes; "X" and "Y" may be any non-Ig moiety.

[0045] Figure 2A shows an lg fusion protein in the Fc-X configuration comprising a hinge from one antibody isotype and an Fc moiety consisting of a CH2 and CH3 domain from a second isotype; at the C-terminus of the Fc moiety is a protein moiety "X".

[0046] Figure 2B shows an lg fusion protein in the X-Fc configuration comprising a hinge from one antibody isotype and an Fc moiety consisting of a CH2 and CH3 domain from a second isotype; at the N-terminus of the Fc moiety is a protein moiety

"X". [0047] Figure 2C shows an lg fusion protein in the X-Fc-Y configuration comprising a hinge from one antibody isotype and an Fc moiety consisting of a CH2 and CH3 domain from a second isotype; at the N-terminus of the Fc moiety is "X", which can be any protein, and at the C-terminus of the Fc moiety is a protein moiety

[0048] Figures 3 A-D are schematic illustrations of lg fusion proteins that comprise variable regions and that contain hybrid isotypes; "X" and "Y" may be any non-Ig moiety.

[0049] Figure 3 A shows an lg fusion protein comprising a hinge from one antibody isotype (black) and CHI, CH2, and CH3 regions from a different isotype; a non-Ig protein "X" is fused at the C-terminus of the heavy chain. [0050] Figure 3B shows an lg fusion protein comprising a hinge from one antibody isotype (black) and CHI, CH2, and CH3 regions from a different isotype; a non-Ig protein "X" is fused at the C-terminus of the heavy chain; arrows indicate a subset of possible sites of mutation in the antibody moiety, as described herein.

[0051] Figure 3C shows an lg fusion protein comprising a hinge and a CHI region from one antibody isotype (black) and CH2, and CH3 regions from a different isotype; a non-Ig protein "X" is fused at the C-terminus of the heavy chain; the branched structure emanating from the hinge represents a glycosylation moiety.

[0052] Figure 3D shows an lg fusion protein comprising a hinge from one antibody isotype (black) and CHI, CH2, and CH3 regions from a different isotype; a non-Ig protein "X" is fused at the C-terminus of the heavy chain; a second non-Ig protein "Y" is fused at the N-terminus of the light chain.

[0053] Figure 3E shows an lg fusion protein comprising a hinge from one antibody isotype (black) and CHI, CH2, and CH3 regions from a different isotype; a non-Ig protein "X" is fused at the C-terminus of the heavy chain; a second non-Ig protein "Y" is fused at the N-terminus of the heavy chain.

[0054] Figure 4 is a schematic illustration of an lg fusion protein that comprises variable regions, comprising multiple isotypes, and that have increased valency compared to IgG; black ovals represent the CH3 and CH4 domains from IgM; white ovals represent the CHI, hinge, and CH2 domains from an IgG; striped ovals represent variable domains and the light chain constant domain; thick lines represent disulfide bonds normally found in IgGl; thin lines labeled with an V represent disulfide bonds normally found in IgM. Definitions

[0055] According to the invention, by isotype is meant the class of heavy chain constant ( C ) region of an immunoglobulin which determines the functional activity of an antibody. There are five major classes including IgA, IgG, IgM, IgD, and IgE. By IgA is meant the class of immunoglobulin characterized by α heavy chains.

By IgD is meant the class of immunoglobulin characterized by δ heavy chains.

By IgE is meant the class of immunoglobulin characterized by ε heavy chains.

By IgM is meant the class of immunoglobulin characterized by μ heavy chains.

By IgG is meant the class of immunoglobulin characterized by γ heavy chains. "Gammal " or "γl " refers to a heavy chain or portion thereof derived from an

IgGl . Similarly, "gamma2" or "γ2" derives from IgG2, and so on. [0056] According to the invention, by allotype is meant allelic polymoφhisms of the same heavy chain C gene of an immunoglobulin. These determinants are found in some, but not all, members of a species. [0057] According to the invention, by idiotype is meant antigenic determinants found on the variable ( N ) regions of antibodies and are due to particular rearranged N_H and V_L genes.

[0058] According to the invention, by FcRn, also know as FcRp, is meant the beta-2 microglobulin-containing neonatal intestinal transport receptor which regulates the clearance of IgG and is important for the in vivo circulating half-life of antibodies.

[0059] According to the invention, by FcγR is meant cell surface receptors, including FcγRI, RII, and RIII, that bind the Fc portion of IgG molecules and elicit effector cell functions. FcγR are expressed on phagocytes, B lymphocytes, ΝK cells and dendritic cells. [0060] According to the invention, by "an IgGl," or "IgG2," or other lg molecule is meant a whole antibody including a heavy and light chain, or a portion thereof.

[0061] According to the invention, by "a bivalent monomer" is meant an antibody, Fc fusion, or antibody fusion that is normally dimerized by the formation of disulfide bonds that form in a normal antibody. The formation of such disulfide bonds is generally infeπed by the ability of an Fc-containing protein to migrate on a denaturing, non-reducing SDS gel as a single band with an apparent molecular weight about twice that of the apparent molecular weight seen with reducing conditions. The presence of a bivalent monomer is also infeπed by the presence of a peak in a size exclusion chromatograph that coπesponds to the coπect molecular weight. Other protein sizing methods may also be used to identify the presence of bivalent monomers.

[0062] According to the invention, by "an lg fusion protein" is meant a fusion protein that contains part or all of an antibody linked to a second moiety which is preferably part or all of a non-antibody or non-immunoglobulin (non-Ig) protein. Immunocytokines, Fc-X proteins, X-Fc proteins, and X-Fc-Y proteins are all examples of lg fusion proteins. Similarly, fusion proteins in which a non-Ig moiety is placed between two lg moieties or domains constitute a type of lg fusion protein.

[0063] According to the invention, by "assembly" of a protein is meant the proper folding of a protein and oligomerization into a coπect multimeric state. Assembly can be monitored in many ways that are established in the art. In practice, coπect assembly of a protein, with respect to disulfide bonds, can be conveniently monitored by comparing migration on a non-reducing and reducing SDS gel: if a given protein species forms multiple bands on a non-reducing gel, but forms a single band on a reducing SDS gel, it can be infeπed that at most only one of the bands on the non- reducing SDS gel is the coπectly assembled species. Alternatively, size exclusion chromatography can be used to distinguish unit proteins from higher-order oligomers and aggregates that may form from incoπect disulfide bonding or non-covalent interactions.

[0064] According to the invention, by "domain" of an antibody moiety is meant a structural domain that coπesponds to the amino acid segments encoded by individual exons in antibody genes in humans. For example, the constant domains of an IgG are the CHI, hinge, CH2, and CH3 domains. In some cases, the hinge and CH2 domains are encoded by the same exon. In such cases, a junction between the hinge and CH2 domain is defined by alignment with other hinge/CH2 junctions (see Paul , op cit., p. 46-49). [0065] According to the invention, by domain, protein, region, or molecule is meant an entire domain, protein, region, or molecule, or a portion, mutant or engineered form thereof, or a form principally derived therefrom. The portion, mutant, or engineered form preferably has functional properties characteristic of the entire domain, protein, region, or molecule. According to the invention, by "principally deriving" is meant having at least 95% of its amino acids deriving from a particular sequence of a naturally occurring protein or domain. For example, a sequence that principally derives from the human IgG2 CH2 domain is at least 95% identical to human IgG2 CH2 domain in an amino acid alignment.

[0066] According to the invention, by "modified IgGl hinge" is meant a hinge region from IgGl in which a cysteine, preferably the first cysteine is mutated to another amino acid. This cysteine normally forms a disulfide bond to the antibody light chain. The modified IgGl hinge is particularly useful in proteins that either lack the light chain or that have another cysteine that can bond to the light chain.

[0067] According to the invention, by "an erythropoietin molecule" is meant a molecule that has generally the same structure and similar amino acid sequence to a vertebrate erythropoietin, optionally including mutations. For example, Example 18 describes the use of human erythropoietin without mutations and of a version of human erythropoietin with four mutations.

Detailed Description of the Invention [0068] The present invention provides methods and compositions for improving the in vivo and in vitro production of immunoglobulin fusion proteins. In particular, the invention provides useful methods for improving the expression, aggregation, and/or folding properties of immunoglobulin fusion proteins. The invention is based in part on the suφrising observation that an immunoglobulin fusion protein is expressed at a higher level, with fewer aggregation and/or folding problems, when a hybrid immunoglobulin is used as the fusion partner instead of a wild-type immunoglobulin. The improved fusion protein production properties associated with hybrid immunoglobulins are unexpected since wild-type immunoglobulins such as IgGl and IgG2 are thought to be well folded proteins that are expressed efficiently both in vivo and in vitro. [0069] Accordingly, one aspect of the invention includes methods and compositions useful for expressing immunoglobulin fusion proteins that include a hybrid immunoglobulin (or hybrid lg) moiety. Prefeπed hybrid immunoglobulins include an IgGl hinge and an IgG2 CH2 and CH3 domains. Other prefeπed hybrids include IgGl and IgG4 domains.

Antibody structure [0070] Antibodies are Y shaped molecules and are composed of two heavy (H) and two light (L) chains. Each heavy chain is linked to a light chain by a disulfide bond and relies on covalent and noncovalent interactions to properly orient the two chains relative to each other. The variable domains at the amino termini of these two chains contain the antigen-binding site and, with the CHI domains, define the Fab end of the molecule.

[0071] The four chains use hydrophobic bonding between the heavy chains and one or more interchain disulfide bonds to stabilize the complex. Thus the complete immunoglobulin is bivalent with two identical antigen-binding sites. Certain immunoglobulins normally undergo additional multimerization, such as IgM and IgA. [0072] Each polypeptide chain has two to five domains; the light chains contain two domains while the heavy chains contain four or five. The single amino terminal domain of each chain is termed variable due to extensive sequence variation, while the several carboxy-terminal domains are refeπed to as the constant regions. Heavy chain regions are numbered CHI, hinge, CH2, CH3, CH4 and are responsible for many important antibody functions including Fc receptor (FcR) binding and complement fixation.

[0073] There are five major isotype classes of heavy chain C regions, classified as IgA, IgG, IgD, IgE, IgM, each with characteristic effector functions (IgG is separated into four γ isotype subclasses: γl, γ2, γ3, γ4). The constant regions of the light chain possess only one C domain and can be one of the two classes, Ck and Cλ, which have no known distinct functional attributes (see, W. E. Paul, ed., 1993, Fundamental Immunology, Raven Press, New York, N.Y).

[0074] All immunoglobulins have a hinge region located C terminal to the CHI domain of their heavy chains separating the Fab and Fc regions of the molecule. In most cases, the hinge regions permit a large degree of flexibility between the antigen binding (Fab end) and effector-interaction (Fc) components of the molecule, thereby linking the two key functional elements of the antibody. In the IgG isotypes, interchain disulfide bonds typically form within this hinge region, generating the final tetrameric molecule.

[0075] Except for IgM, the hinge region is dominated by prolines, serines and threonines, amino acids that tend to prevent formation of secondary structures and that are thought to give hinges flexibility. The IgGl isotype, often used for fusion proteins, has two disulfide bonds in the hinge region that link the heavy chains to one another. In contrast, the IgG2 isotype has four disulfide bonds (Figure 1). According to the invention, these disulfide bonds tend to promote incoπect assembly of antibodies and lg fusion proteins, as shown in the unexpected findings in the Examples.

Useful configurations of lg fusion proteins

[0076] Immunocytokines are only one example of a tumor-targeted fusion protein therapy for which methods and compositions of the invention are useful. Other tumor-toxic molecules may also be targeted to tumors by fusion to tumor-specific antibodies according to the invention. In addition, other types of diseased cells, such as virus-infected cells, may be attacked by antibody fusion proteins according to the invention.

[0077] Methods and compositions of the invention are also useful in combination with the use of Fc-X and X-Fc technology. According to the invention, the use of hybrid antibody isotypes in a fusion protein further improves the production and collection of a target protein or polypeptide of interest that is linked to the Fc portion of an immunoglobulin. For Fc-X fusion proteins, a signal peptide, followed by the Fc fragment of an immunoglobulin gene is the N-terminal fusion partner to the target protein. The fusion protein is then expressed in a host cell, for example a mammalian cell such as a cell that naturally expresses the immunoglobulin. The signal peptide-Fc fragment in the a N-terminal fusion partner directs the target protein through the secretory pathway such that the fusion protein is readily secreted. Additionally, use of the Fc fragment, which is normally glycosylated and highly charged at neutral pH, facilitates solubilization of more hydrophobic proteins. The targeting of Fc-X fusion proteins through the secretory pathway also alleviates problems associated with protein toxicity inside the cell and facilitates the isolation of stable cell lines. The fusion protein product is easily assayed and purified as it is readily collected from the culture media in its native conformation with retention of both biological and enzymatic activities. The efficacy of this technology has been demonstrated with Fc- leptin and Fc-erythropoietin. Some of these advantages are also inherent in X-Fc proteins. [0078] According to the invention, examples of useful fusions of antibody moieties include Fc-X, X-Fc, and X-Fc-Y proteins. Such proteins contain an Fc region with a non-antibody protein or protein fragment fused at the N-terminus, the C- terminus, or both the N-terminus and C-terminus. One advantageous effect of fusion to the Fc region is that the serum half-life of the fusion partner can be significantly extended. A second advantageous effect is that an 'X' can be effectively dimerized by attachment to the Fc. For example, Enbrel is a fusion protein consisting of a portion of a TNF receptor and a human IgGl Fc region.

[0079] In some embodiments of the invention, it is specifically advantageous to engineer a fusion protein with hybrid isotypes in the X-Fc orientation. With these constructs the target protein is the N-terminal fusion protein and the Fc fragment follows. For some proteins this approach may be useful, for example with lymphocyte cell surface glycoprotein (LHR) (see US patent 5,428,130).

[0080] According to the invention, the utility of recombinant antibody-based fusion proteins, albeit better than protein or cytokine treatments alone, may be limited by their rapid in vivo clearance from the circulation since antibody fusion proteins have a significantly lower in vivo circulating half-life than the free antibody. The decreased circulating half-life is likely a result of increased clearance via the Fc receptor (FcR). Switching of isotype has proven to be one mechanism by which features such as half-life can be altered. Improvements in the half- lives of two immunocytokines have been demonstrated (Cancer Research 59(9):2159-66, 1999) by changing the isotype of the human heavy chain C region from IgGγl or IgG γ3 to IgG γ4, an isotype with reduced FcR binding. The IgG4-based immunocytokines and fusion proteins have 10- fold reduced FcR binding and reduced Fc receptor effector functions such as ADCC, but still show similar or better efficacy in mouse tumor models than that of the original IgGγl based fusion proteins. However, the invention provides fusion proteins based on hybrid antibodies that combine functional and structured properties of different antibody types in a single molecule. [0081] Accordingly, for certain applications, the IgG2 isotype confers superior qualities to the antibody fusion protein since the IgG2 isotype has greatly reduced Fc receptor binding (Hulett et al. [1994] Adv. Immunol.57: 1127). As with whole antibody fusion proteins, it is sometimes advantageous to use a γ2 isotype in fusion proteins containing only an Fc region. The rationale is the same as for whole antibodies: it is often desirable to avoid the binding to Fc receptors that is mediated by Fc regions derived from other isotypes.

[0082] However, the use of the IgG2 isotype in a fusion protein generally causes some level of inappropriate assembly, as described herein. Methods and compositions of the invention provide antibody fusion proteins that have the minimized effector functions of the IgG2 isotype, but which do not have the aggregation properties of this isotype.

[0083] According to the invention, novel hybrid isotype antibodies and fusion proteins show enhanced expression and improved assembly over that of IgG2-based fusion proteins. Additionally, hybrid isotype antibodies and fusion proteins may have increased efficacy in therapeutics where Fc receptor effector functions are not desired.

Types of Hinges

[0084] Hybrid isotype antibodies of the invention also include antibodies wherein the hinge region is a mutant hinge region, preferably a hinge region with a reduced number of cysteine residues, for example, a modified IgGl hinge region in which the first cysteine is mutated to serine. The first cysteine of the IgGl hinge region normally bonds to the light chain. In an Fc-X protein or an X-Fc protein or any other antibody fusion protein lacking a light chain, this cysteine does not serve its natural function and therefore can be mutated. However, according to the invention, a hybrid isotype antibody with an IgGl hinge missing the first cysteine and an IgG2 CHI domain can associate with a light chain, because the light chain normally forms a disulfide bond with a cysteine within the IgG2 CHI domain. The four cysteines within the IgG2 hinge are thought to form disulfide bonds with each other. [0085] According to the invention, the cysteines in a hinge region of an antibody or lg fusion protein that are involved in heavy chain-heavy chain homodimerization can have significant effects on the expression and assembly of antibodies or lg fusion proteins. In particular, a higher number of cysteines can lead to incoπect assembly of an antibody or lg fusion protein, due to incoπect disulfide bond formation. Thus, the invention discloses that mutation of one or more cysteines involved in heavy chain homodimerization can cause an improvement in expression or assembly of an antibody or lg fusion protein. In a prefeπed embodiment, a heavy chain- homodimerization cysteine can be mutated to, in order of general preference, serine, alanine, threonine, proline, glutamic acid, glutamine, lysine, histidine, arginine, asparagine, aspartic acid, glycine, methionine, valine, isoleucine, leucine, tyrosine, phenylalanine, tryptophan or selenocysteine. [0086] It is particularly convenient to use an IgGl hinge that is mutated in the first, most N-terminal cysteine. The advantage of this hinge region is that it only has two cysteines. The first cysteine in the IgGl hinge normally forms a disulfide bond with the a cysteine in the light chain. However, in lg fusion proteins lacking a light chain, such as Fc-X, X-Fc and X-Fc-Y proteins, the most N-terminal cysteine in the IgGl hinge serves no such puφose and therefore can be mutated (Lo et al. Protein Engineering 11:405-500 [1998]). As described in the Examples, the two-cysteine IgGl hinge can also be used in an intact antibody or intact antibody fusion protein in which the CHI domain derives from IgG2, since the IgG2 CHI domain has a cysteine that can form a disulfide with the light chain. Fc Receptors

[0087] IgG molecules interact with multiple classes of cellular receptors including three classes of Fcγ receptors (FcγR) specific for the IgG class of antibody, namely FcγRI, FcγRII, and FcγRIII. These receptors are responsible for uptake of antigen- antibody complexes. The Fc receptor binding site on Fc is found on the CH2 domain near the hinge, and it is thought that binding of the V regions to an antigen helps displace the light chain constant region from sterically blocking the hinge. In this way, antibodies bound to antigen are preferentially bound by Fc receptors.

[0088] A fourth receptor, alternatively termed the 'protection receptor' (FcRp) or the 'neonatal receptor' (FcRn), is responsible for recycling antibodies from the endosome after the endocytosis of antibody-antigen complexes and their disaggregation in the endosome. The binding site for FcRp is found at the junction between the CH2 and CH3 domains in the three-dimensional antibody structure. The serum half-life of IgG antibodies depends on a productive interaction with functional FcRp. Other antibody types, such as IgM, IgD, IgE, and IgA do not bind to FcRp.

[0089] Another binding partner of certain antibodies is C 1 q, which mediates complement fixation. [0090] Interaction with Fc receptors and FcRp also affects the biological activity and metabolism of fusion proteins containing an Fc moiety.

[0091 ] For example, fusion proteins with poor binding to FcR have longer serum half-lives than coπesponding fusion proteins with good binding to FcR. Fusion proteins with poor binding to FcRp have shorter serum half-lives than coπesponding fusion proteins with good binding to FcRp.

[0092] For example, lg fusion proteins containing the CH2 and CH3 domains of IgG2 have a longer serum half-life than fusion proteins containing an IgGl . Similarly, fusion proteins containing an Fc region derived from IgG3 have a shorter serum half-life than coπesponding fusion proteins containing IgGl or IgG2. Fusion proteins containing CH2 and CH3 domains derived from IgM, IgD, IgE, and IgA have even shorter serum half-lives than coπesponding IgG-derived fusion proteins.

[0093] To illustrate the uses of the invention, it is convenient to group antibodies and lg fusion proteins into two general classes: proteins in which immunological effector functions are desired and proteins in which the lg moiety serves as an immunologically inert carrier and lacks effector functions. In the former category of proteins it is convenient to construct proteins with hybrid isotypes in which a particular constellation of effector functions is created. In the latter category it is convenient to construct proteins with hybrid isotypes involving regions of certain isotypes with minimal effector functions and regions from other isotypes that enhance the assembly of the protein as described below.

Assembly of antibodies and lg fusion proteins

[0094] The invention discloses the discovery that the hinge of an antibody or an lg fusion protein plays a key role in the proper assembly and lack of aggregation of a fusion protein, such as an lg fusion protein secreted from a mammalian cell. For example, without wishing to be bound by theory, it is thought that antibody and lg fusion protein assembly involves a step in which two heavy chains first non- covalently associate by a hydrophobic patch in the CH3 domain. After this association, the hinge regions are aligned and interchain disulfide bonds are formed. The hinge region is about 50 Angstroms from the hydrophobic patch in the CH3 domain. [0095] In designing an antibody or lg fusion protein construct, it is useful to vary the hinge region. For example, it is useful to replace the DNA encoding a hinge region in a given expression construct with DNA encoding a different hinge region, such as the hinge region from IgGl, IgG2, IgG3, IgG4, IgM, IgA, IgD, IgE, or IgY.

[0096] According to a theory of the invention, for example, antibodies and lg fusion proteins containing hinge regions with a greater number of cysteines do not assemble as efficiently as coπesponding proteins with fewer cysteines. Without wishing to be bound by theory, hinge regions with larger number of cysteines exhibit a greater chance for misalignment of the cysteines and incoπect disulfide bond formation. As a result, antibodies and lg fusion proteins with hinge regions containing a larger number of cysteines are found to be more highly aggregated and exist in multiple electrophoretic species as compared to coπesponding antibodies and lg fusion proteins with hinge regions containing a smaller number of cysteines. Illustrations of this phenomenon are given in the Examples.

[0097] For example, antibodies and lg fusion proteins containing a hinge region containing four cysteines are less efficiently assembled than antibodies and lg fusion proteins containing three or two cysteines. For example, proteins containing an IgG2- derived hinge region are less efficiently assembled than coπesponding proteins containing an IgGl -derived hinge region. In a distinct example, proteins containing an IgG3 -derived hinge region are poorly assembled; the IgG3 hinge region contains 11 cysteines.

[0098] The utility of the invention is particularly illustrated in situations in which minimal binding to an Fc receptor I is desired. For example, in the context of fusion proteins, it is convenient to use the CH2 and CH3 regions from IgG2 because binding to FcR is profoundly reduced, so that ADCC is reduced and serum half-life is enhanced. However, use of the IgG2 hinge region causes the resultant antibody or lg fusion to be poorly assembled. It is particularly convenient to use an IgG2 CH2 and CH3 region in combination with an IgGl hinge, as illustrated in various Examples below.

[0099] A parallel finding, illustrated in several Examples, is that the choice of a particular hinge or other domain may affect the production levels of an antibody or lg fusion. This finding is of great economic importance, since protein drugs such as intact antibodies often need to be dosed in large amounts, such as several hundred milligrams per patient per dose. It is generally found that the choice of a hinge with a minimal number of cysteines, such as three or two, improves the yield of an antibody or lg fusion protein from a eukaryotic expression system. Lowering the number of cysteines may be accomplished by mutation or by substituting the hinge of one isotype for another, or both.

Enhancement of protease resistance

[0100] The hinge region is particularly sensitive to proteases. In classical experiments, antibodies were divided into Fab regions and Fc regions by protease cleavage in the hinge region.

[0101] It is convenient to construct antibodies and lg fusion proteins with a hinge region from IgA and other components from other antibody isotypes. For example, if an antibody or lg fusion protein is to be delivered orally or across another mucosal surface such as the nasal, lung, vaginal, or rectal mucosa, it is useful to have a particularly protease-resistant protein to protect the antibody or lg fusion protein from proteases that are present. For example, it is useful to construct an antibody or lg fusion protein that contains an IgA hinge and the CH2 and CH3 domains from an IgG, so that this hybrid isotype protein will have the features of protease resistance and an extended serum half-life once the protein enters the circulation. The hinge of IgA 1 is a prefeπed hinge, because the glycosylation sites in the IgAl hinge confer resistance to a broad spectrum of proteases. In contrast, the hinge of IgA2 is shorter and is resistant to bacterial proteases that specifically cleave the IgAl hinge.

[0102] Other isotype heavy chains also contain glycosylation sites that contribute to protease resistance. For example, IgA, IgD, IgE, and IgM contain glycosylation sites in the constant domains that contribute to protease-resistance. For example, the IgE CHI domain contains three N-linked glycosylation sites. It is useful to combine the IgE CHI domain with, for example, a hinge region from IgA and with CH2 and CH3 domains from an IgG.

[0103] It is also useful to incoφorate glycosylation sites from one isotype into the CH2 or CH3 region of another isotype. In such cases, it is generally not sufficient to construct antibodies or lg fusion proteins in which entire domains, as defined by regions encoded by single exons, and so it is useful to construct hybrid domains. For example, it is useful to combine the features of FcRp binding characteristic of IgG isotypes and the protease-resistance features of other isotypes. To achieve such a combination of features, it is necessary to construct individual domains with amino acid sequences from two different isotypes. The resulting hybrid domains are then used to construct lg fusions to non-Ig moieties.

[0104] For example, it is convenient to substitute a stretch of amino acids from the IgE CH2 domain that include the sequence NNLTW (SEQ ID NO: 3) for the coπesponding amino acids in an IgG CH2 domain. For example, in IgGl these amino acids are VKFNW (SEQ ID NO: 4) and in IgG3 these amino acids are NQFKW (SEQ ED NO: 5). According to the invention, the coπesponding amino acids in other CH2 domains can be determined by computer-based or structural alignment of the CH2 domains of other isotypes, or from the published literature (Paul, WE Fundamental Immunology Fourth Edition, chapter 3, p. 46-47). [0105] Similarly, it is also useful to incoφorate other glycosylation sites into IgG. For example, sequence segments that include the NTSGF (SEQ ID NO: 6) or LNASR (SEQ ID NO: 7) sequences from IgD are used to replace the coπesponding sequences of an IgG-derived Fc region in an antibody or lg fusion protein. According to the invention, other useful glycosylation sites within antibody constant regions are disclosed in Paul (Fundamental Immunology Fourth Edition, chapter 3) and references therein.

[0106] In some embodiments of the invention, the incoφoration of a non-IgG glycosylation site reduces the interaction with FcRp. In such cases, there is a tradeoff between the improvement in protease resistance and the decrease in serum half-life, and the usefulness of such a hybrid isotype protein must be evaluated in the context of a particular application. [0107] According to the invention, the protease resistance of a particular hybrid isotype antibody or lg fusion protein is assayed according to standard procedures. Proteases can be purchased from commercial suppliers and are incubated in the presence of a particular hybrid isotype antibody or lg fusion protein in accordance with the manufacturer's specifications. Proteolysis is measured, for example, by SDS gel electrophoresis and quantitation of starting material and cleavage products.

[0108] According to the invention, the ability of a hybrid isotype antibody or lg fusion protein to interact with FcRp is also measured according to standard procedures. For example, the pharmacokinetic properties of an antibody or lg fusion protein are measured in a mammal such as a mouse, rat, rabbit, dog, non-human primate, or human. The pharmacokinetic properties of an antibody or lg fusion protein are a practical indication of FcRp binding, and in general, the p pose of incoφorating the feature of FcRp binding into an antibody or lg fusion protein is to improve the pharmacokinetic properties of an antibody. It is also convenient to examine the three-dimensional structure of the FcRp-Fc complex to determine whether a particular hybrid isotype antibody or lg fusion protein is likely to interact with FcRp (Martin, W. L., et al.: Crystal Structure at 2.8A of an FcRn/Heterodimeric Fc Complex: Mechanism of pH-Dependent Binding. Mol. Cell 7 pp. 867 [2001]; structure ID 111 A at http://www.rcsb.org/pdb . [0109] According to this aspect of the invention, it is particularly useful to construct proteins of the form Fc-X, wherein the Fc region contains a hinge region from IgAl, a CH2 and CH3 region containing those elements of IgG2 that mediate binding to FcRp, and those elements of CH2 that have reduced effector functions compared to other IgGs.

Increasing the valency of an antibody or lg fusion protein

[0110] According to the invention, it is sometimes useful to construct a hybrid isotype antibody or lg fusion protein that has a high valency, but that also has effector functions or other features characteristic of lower-valency antibodies. IgA and IgM have a high valency due to oligomerization via interchain disulfide bonds in the CH3 region. IgA and IgM also have disulfide bonds from a cysteine near the C-terminus of CH4 to the J chain. IgA is dimeric and IgM is pentameric or hexameric, such that IgA has 4 antigen-binding sites and IgM has 10 or 12 antigen-binding sites. [0111] However, IgM and IgA do not mediate ADCC. To construct a polyvalent antibody or lg fusion that mediates -ADCC, it is useful to construct a protein that has the CH2 domain of an IgG and the CH3 and CH4 domains of IgM or IgA. In general, it is preferable to use IgM CH3 and CH4 domains, since the resulting hybrid isotype protein has a higher valency than the coπesponding IgA-containing hybrid isotype protein.

[0112] The following application illustrates the utility of hybrid isotype fusion proteins with increased valency. Many tumor cells overexpress the EGF-receptor on their cell surfaces. The EGF-receptor is also expressed on the surface of many normal cells, so the difference in EGF-receptor expression between normal and tumor cells is only quantitative. According to one embodiment of the invention, a useful IgG-IgM hybrid isotype protein comprises a V region that interacts weakly with the human EGF-receptor. The affinity of the N region is chosen so that the fusion protein does not efficiently interact with EGF-R on normal cells, but interacts with the overexpressed EGF-receptor on tumor cells as a result of an avidity effect. The use of a CH2 domain of an IgG, for example the CH2 domain of IgGl, mediates ADCC against the tumor cells.

[0113] To enhance the specific killing of tumor cells, it is also useful to fuse the anti-EGF-R IgG-IgM hybrid isotype protein to a protein with anti-tumor activity, such as a cytokine. For example, IL-2 may be used. Alternatively, it is useful to conjugate a radioactive atom to the hybrid isotype protein, such that concentration of the hybrid isotype protein in the region of the tumor causes preferential iπadiation of the tumor. For example, Yttrium-90 may be conjugated to the IgG-IgM hybrid isotype protein. The combination of ADCC and IL-2 action or ADCC and iπadiation is particularly useful in killing tumor cells.

[0114] In an IgG-IgM fusion protein, such as the anti-tumor protein described above, it is also generally useful to use a hinge region from IgG, preferably IgGl. The hinge region from IgG3 is, for many applications, the least prefeπed IgG hinge region, because this hinge region tends to cause variable assembly; moreover, the IgG3 hinge is easily proteolyzed (Baici A, et al., Scand J Immunol. [1980] 12:41-50).

[0115] The preceding example illustrates a general principle of this aspect of the invention: that an antigen that is more highly expressed on a target cell type than on normal cells may be more effectively targeted by an antibody or lg fusion protein with a high valency but a relatively low monovalent affinity.

[0116] For example, in autoimmune diseases, certain immune cells such as T cells express a higher level of cell surface proteins such as cytokine receptors. It is useful to attack such immune cells with a high valency IgGl, IgG2, or IgG4-IgM antibody or lg fusion protein directed against the up-regulated surface protein. In this way, targeting of cells in which the cell surface protein is present but not up-regulated is minimized. For example, during an attack of an autoimmune disease, it is useful to treat a patient with an IgG-IgM fusion protein in which the V regions are directed against, for example the IL-2 receptor, the IL-12 receptor, or any other up-regulated receptor. The hinge region and the CH2 domain are preferably derived from IgGl, IgG2, or IgG4, and the CH3 and CH4 regions are preferably derived from IgM. The N regions preferably bind weakly to the IL-2 or IL-12 receptor for the treatment of autoimmune disease. This treatment has the effect of killing a subset of T cells but not the entire T cell repertoire.

Construction of expression plasmids that express antibodies and lg fusion proteins with hybrid isotypes.

[0117] The invention also provides nucleic acids that encode the antibodies and fusion proteins of the invention. The invention is best practiced when the encoding nucleic acids also encode a secretion cassette that, when transcribed and translated, gives rise to a signal peptide. The signal peptide is generally cleaved from the mature product. The secreted fusion protein can be collected from the culture media without the need for lysis of the host cell, and can be assayed for activity or purified using common reagents as desired. In some cases, the presence of certain fusion partners, such as the cytokine CΝTF, allows for secretion of an lg fusion protein without a secretion cassette.

[0118] One of ordinary skill in the art can perform such recombinant DΝA constructions using DΝA with introns, because the naturally occuring introns separate the DΝA encoding the hinge from the DΝA encoding the CHI and CH2 domains. Restriction sites in the introns can be used. Alternatively, because the hinge regions are generally only about 15 to 70 amino acids long, it is possible to construct synthetic DΝAs encoding an entire hinge region, for example with oligonucleotide synthesis, PCR, or a combination thereof. The synthetic hinge-encoding region can then be placed into an expression plasmid encoding an lg fusion protein using standard recombinant DNA techniques.

[0119] The invention is illustrated further by the following non-limiting examples.

EXAMPLES Example 1 : Construction of a plasmid expressing an Fc-X fusion protein with a hinge region and CH2 region from different antibody isotypes. [0120] The construction of a plasmid that expresses HuFcγl -Leptin has been described in the PCT publication WO00/040615 A2.

[0121] A plasmid that expresses a fusion of an IgG2-derived Fc and a C-terminal fusion partner was constructed as follows.

[0122] First, the genomic sequence of human γ2 Fc was obtained. The genomic DNA encoding human Fcγ2 was obtained by PCR on cellular DNA isolated from human PBMCs. The forward primer had the sequence 5 ' CC TTA AGC GAG CGC AAA TGT TGT GTC GAG (SEQ ID NO: 8), where an Aflll restriction site C TTA AG was introduced just upstream of the γ2 hinge coding region GAG CGC AAA TGT TGT GTC GAG (SEQ ID NO:9). The reverse primer had the sequence 5 ' CCTCGAG TCA TTT ACC CGG GGA CAG GGA G (SEQ ID NO: 10), where an Xhol restriction site CTCGAG was introduced immediately after the translation stop codon (anticodon TCA). In addition, the reverse primer also introduced a Smal CC CGGG by silent mutation (A to G substitution underlined). The 910 bp PCR fragment was cloned into the TOPO TA Cloning Vector (Invitrogen, Carlsbad, CA) for sequence verification. [0123] Second, the human γ2 Fc was placed into an expression vector. The natural Smal restriction site in the DNA sequence CTG CCC CCA TCC CGG GAG GAG ATG ACC AAG (SEQ ID NO: 11) encoding the upper CH3 region was deleted by a silent mutation introduced by an overlapping PCR technique (Daugherty, B.L. et al., Nucleic Acids Res. 19:2471-6, 1991). The forward primer had the sequence 5' CTG CCC CCA TCA CGG GAG GAG ATG ACC AAG (SEQ ID NO: 12), where the C to A substitution is underlined; and the reverse primer had the sequence 5' GGT CAT CTC CTC CCG TGATGG GGG CAG GGT GTA C (SEQ ID NO: 13), where the G to T substitution is underlined. After sequence verification, the resultant Aflll- Xhol restriction fragment encoding the Fc of γ2 contained a unique Smal site upstream of the translation stop codon, and a Xhol site downstream of the stop codon. The Aflll-Smal fragment encoding Fcγ2 was then used to replace the coπesponding restriction fragment encoding Fcγl in pdCs-huFcγl -Leptin (PCT publication WO00/040615A2) to give pdCs-huFcγ2-Leptin.

[0124] Third, the DNA encoding the hinge of the Fc γ2 was replaced by the altered hinge from γl. The γ2 hinge region contains four cysteine disulphide bonds. Shown below is the Aflll-Stul fragment containing the γ2 hinge exon. The Aflll site (C TTA AG) is preceded by DNA sequence encoding a signal peptide. Glutamic acid (E) is the first amino acid residue of the γ2 hinge. The letters in small case represent the intron sequence following the γ2 hinge exon. The Stul restriction site (aggcct) is C-methylated in most E. coli strains by the DCM methylase because of the C- methylation in the sequence ccagg and the reverse strand cctgg; when methylated, this site cannot be cut by Stul.

E R K C C V E C P P C P (SEQ ID NO: 14)

C TTA AGC GAG CGC AAA TGT TGT GTC GAG TGC CCA CCG TGC CCA G gtaagccagcccαggcct (SEQ ID NO: 15)

[0125] The Aflll-Stul fragment containing the γ2 hinge exon in pdCs-huFcγ2- Leptin was replaced by the coπesponding Aflll-Stul fragment containing the γl hinge exon from pdCs-huFc γl -Leptin, which is shown below:

E P K S _.S D K T H T C P P C P (SEQ ID NO: 16)

C TTA AGC GAG CCC AAA TCT TCT GAC AAA ACT CAC ACA TGC CCA CCG TGC CCA G Gtaagccagcccαggcct (SEQ ID NO: 17) [0126] The γl hinge sequence in pdCs-huFc γl contains a Cys to Ser mutation (underlined) which eliminates the Cys residue that forms a disulphide bond with the light chain in IgGl (Lo et al., (1998) Protein Engineering 11 :495-500). Since the Stul sites in both the γl and γ2 exons are C-methylated and the Stul restriction endonuclease is methylation sensitive, both plasmids were isolated from a DCM negative strain of bacteria before digestion with the Stul enzyme. The resultant pdCs- huFcγ2-Leptin with the hinge region from pdCs-huFc γl was designated pdCs- huFcγ2h-Leptin (γ2h: γ2 with altered hinge).

Example 2: Characterization of the oligomerization state of huFc γ2-leptin and huFcγ2h-leptin immunofusion.

[0127] Protein expression from NS/0 cells using expression vector pdCs-huFc- huLeptin with the γl, γ2, and γ2h isotypes was evaluated. The physical state of different forms of huFc-huLeptin in which the huFc moiety derived from γl, γ2, and γ2h isotypes were evaluated.

[0128] DNA constructs generated as described above and as in Lo et al. were transfected into NS/0 cells and stable expressing cell lines were generated according to standard procedures.

[0129] In this and the following Examples, stable transfectants were generated as follows. Plasmid DNAs were introduced into the mouse myeloma NS/0 cells by electroporation. NS/0 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 2 mM glutamine and penicillin/streptomycin. About 5xl0⁶ cells were washed once with PBS and resuspended in 0.5 ml PBS. Ten μg of linearized plasmid DNA were then incubated with the cells in a Gene Pulser Cuvette (0.4 cm electrode gap, BioRad) on ice for 10 min. Electroporation was performed using a Gene Pulser (BioRad, Hercules, CA) with settings at 0.25 N and 500 μF. Cells were allowed to recover for 10 min. on ice, after which they were resuspended in growth medium and then plated onto two 96 well plates. Stably transfected clones were selected by growth in the presence of 100 nM methotrexate (MTX), which was introduced two days post-transfection. The cells were fed every 3 days for two to three more times, and MTX-resistant clones appeared in 2 to 3 weeks. Supernatants from clones were assayed by anti-Fc ELISA to identify high producers. High producing clones were isolated and propagated in growth medium containing 100 nM MTX.

[0130] Concentration of HuFc-huLeptins in the supernatant was determined by an anti-huFc ELISA, using anti-huFc antibody (horseradish peroxidase-conjugated goat anti-huIgG, Fcγl or Fcγ2, from Jackson ImmunoResearch). Relatively low levels of expression were detected in the supernatants of the γ2 constructs whereas the γl and γ2h constructs gave high levels of expression in both transient and stable transfections. In transient transfections with equivalent expression vectors encoding huFcγ2-huLeptin and huFcγ2h-huLeptin, the huFcγ2h-huLeptin was produced at 8- fold higher levels.

[0131] For purification, the fusion proteins in tissue culture supernatants were bound to Protein A Sepharose followed by elution in a sodium phosphate buffer (100 mM NaH₂PO₄, pH 3, and 150 mM NaCl). The eluate was then neutralized with 0.1 volume of 2 M Tris-hydrochloride, pH 8 and evaluated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and HPLC - size exclusion chromatography (SEC).

[0132] Fusion proteins were examined by SDS-PAGE under non-reducing conditions. The protein bands, visualized by Coomassie staining, showed HuFcγl- huLeptin to have an apparent MW of 96 kD, indicating a bivalent monomer. Analysis by SDS-PAGE showed that much of the huFcγ2-huLeptin fusion protein was in a higher molecular weight form, migrating as a ladder of bands with apparent molecular weights much higher than Fc-Leptin. In contrast, huFcγ2h-huLeptin was primarily a single species, migrating at 96 kD.

[0133] Size exclusion chromatography (SEC) analysis of huFcγ2h construct coπelated with SDS-PAGE results and showed both huFcγ2h-leptin and huFcγl- leptin to be about 83% monomeric, while the comparable huFcγ2-huLeptin fusion protein was about 55 % monomeric.

[0134] Studies with fixed J774 cells, which are rich in the FcγR class of receptors, showed that only the huFcγl-huLeptin fusion protein exhibits Fc binding. In addition, BAF-3 cells transfected to express the leptin receptor so that their proliferation could be stimulated by leptin (Gainsford, T., et al. PNAS [1996] 93:14564-14568) were used. Studies with BAF3/leptin receptor cells showed that huFcγl-huLeptin, huFcγ2- huLeptin and huFcγ2h-huLeptin were equivalent in leptin bioactivity.

[0135] Stable clones of mammalian cells expressing huFcγ2-huLeptin and huFcγ2h-huLeptin were also identified. Cell transfections were carried out under essentially identical conditions and an identical number of stably transfected cells were cloned and tested for production of huFcγ2-huLeptin and huFcγ2h-huLeptin. The best-expressing clone for huFcγ2h-huLeptin made about 5-fold more huFc- huLeptin than the best-expressing clone for huFcγ2-huLeptin.

Example 3: Construction of a plasmid expressing an X-Fc fusion protein with a hinge region and CH2 region from different antibody isotypes.

[0136] The synthetic DNA sequence (SEQ ID NO: 18) encoding glucagon-like peptide 1 (GLP-1) amino acid residues 7 to 37 (SEQ ID NO: 19) is disclosed below.

H A E G T F T S D V S S Y L E G

C TTA AGC CAT GCT GAA GGG ACC TTT ACT AGT GAT GTA AGT TCT TAT TTG GAA GGC

Q A A K E F I A W L V K G R G CAA GCT GCC AAG GAA TTC ATT GCT TGG CTG GTG AAA GGC CGA GGA GGA TCC TTA

AGC

[0137] The DNA encoding the GLP-1 peptide was preceded by C TTA AGC, where the Aflll restriction site was used to join this DNA fragment to the DNA fragment encoding the signal sequence (Lo et al. Protein Engineering). At the 3' end, the DNA encoding GLP-1 was followed by a BamHI restriction site (GGA TCC, which encodes the amino acid residues G and S), and an Aflll restriction site, which was used to ligate to the Aflll-Xhol restriction fragment encoding Fc γ 2 (or Fc γ 2h) with a translation stop codon (see Example 1).

Example 4: Characterization of the oligomerization state of GLPl-huFc γ2 and GLPl-huFc γ2h immunofusions.

[0138] The resultant vectors from Example 3 , pdCs GLP- 1 huFcγ 1 , pdCs GLP- 1 huFcγ2, and pdCs GLP-1 huFcγ2h, were used to transiently and stably transfect mammalian cells for the expression of GLP-1 (7-37) huFc γl, GLP-1 (7-37) huFc γ2 and GLP-1 (7-37) huFc γ2h.

[0139] Assessment of aggregation state and total protein expression for each GLP-1 huFc fusion protein was performed by SDS-PAGE and HPLC-SEC after purification by Protein A Sepharose, using the general methods described in Example 2. The protein bands were visualized by Coomassie staining. GLP-1 huFcγ2 and GLP-1 huFcγ2h had an apparent MW of approximately 60 kD by SDS-PAGE. Concentration of GLP-1 huFc variants in the supernatant was determined by anti- huFc ELISA. In transient transfections with equivalent expression vectors encoding GLPl-huFcγ2 and GLPl-huFcγ2h, the GLPl-huFcγ2h was produced at about 1.5-fold higher levels.

[0140] Analysis of total cell lysates by SDS-PAGE showed that about half of the GLPl-huFcγ2 fusion protein had incoπect disulfide bonds, as illustrated by the presence of several high molecular weight forms migrating with an apparent weight of 100 to 200 kD. In contrast, essentially all of the detectable GLPl-huFcγ2h fusion protein migrated with an apparent molecular weight of about 60 kD. When the samples were reduced before SDS-PAGE, the GLPl-huFcγ2 and GLPl-huFcγ2h proteins ran with an essentially identical apparent molecular weight of about 34 kD.

[0141] Comparative analysis by HPLC-SEC of the γl, γ2 and γ2h fusion proteins demonstrated that the modified γ2h fusion protein was significantly more monomeric than the other fusion proteins. The modified GLPl-huFcγ2h fusion protein, as illustrated by a single peak, was 84% bivalent monomer, while both the GLPl-huFcγl and GLPl-huFcγ2 fusions proteins had multiple peaks coπesponding to approximately 42 and 33% bivalent monomer, respectively. [0142] Thus, the GLP-l-Fc fusion protein bearing the hinge of IgGl and the CH2 and CH3 domains of IgG2 showed smprisingly better assembly properties than GLP- l-Fc fusion proteins in which the entire Fc region derived from either IgGl or IgG2. Example 5: Construction of an intact antibody specific for diseased cells that contains IgG2 CHI. CH2. and CH3 domains and an IgGl hinge. [0143] The genomic DNA encoding the immunoglobulin γ 2 constant regions (CHI, hinge, CH2 and CH3) was obtained by PCR using cellular DNA isolated from human PBMCs. The forward primer had the sequence 5' CA

AGCTTTCΥGGGGCGAGC (SEQ ID NO: 20), where a Hindlll restriction site A AGCTTv/as introduced to the intron sequence about 220 bp upstream of the CHI exon. The reverse primer has the sequence 5' CCTCGAG TCA TTT ACC CGG GGA CAG GGA G (SEQ ID NO: 21), where an Xhol restriction site CTCGAG was introduced immediately after the translation stop codon (anticodon TCA), and a Smal CC CGGG was created by silent mutation, as already described in Example 1. The natural Smal restriction site in the DNA sequence encoding the upper CH3 region was also deleted by a silent mutation introduced by overlapping PCR, as described in Example 1. One skilled in the art would also recognize that by taking advantage of restriction fragments encoding the Fcγ2 obtained in Example 1 , a ~1810 base pair (bp) Hindlll-Xhol restriction fragment encoding the CHI, hinge, CH2 and CH3 regions and containing a unique Smal restriction site could be readily constructed. After sequence verification, the Hindlll-Xhol fragment encoding γ2 constant regions was used to replace the Hindlll-Xhol fragment encoding γl constant regions-IL2 in pdHL7-huKSγl -IL2 to give pdHL7-huKSγ2 antibody.

[0144] The expression vector for huKSγ2h antibody was constructed as follows. The plasmid pdCs-huFcγ2h-Leptin was used as the PCR template for creating a -130 bp Pstl-PvuII restriction fragment encoding the cysteine-to-serine modified γl hinge region only. The forward primer had the sequence 5' CTGCA GAGCCCAAATCTTC (SEQ ID NO: 22), which restored the native Pstl (CTGCAG) restriction site at the beginning of the γl hinge exon (with the C to S mutation described above). The reverse primer had the sequence 5' CA GCTGGGGCCTGTCCCTG (SEQ ID NO: 23), which coπesponded to the PvuII site (CAGCTG) in the intron between the hinge and the CH2 exons. After sequence verification, this -130 bp Pstl-PvuII restriction fragment was used to replace the coπesponding fragment in pdHL7- huKSγ2 antibody to give the pdHL7-huKSγ2h antibody. Example 6: Characterization of the non-reduced state of a γ2-antibody and a coπesponding γ2h antibody directed against diseased cells.

[0145] For transient transfection, the vectors, pdHL7 of KS antibody with IgGγl, γ2, and γ2h isotypes were introduced into mammalian cells by lipofection using Lipofectamine Plus (Life Technologies, Gaithersburg, MD) according to supplier's protocol. Stable transfectants were generated as described in Example 2.

[0146] KS antibodies in the conditioned media (10% serum) were captured on Protein A Sepharose (Repligen, Cambridge, MA) and then eluted by boiling in the protein sample buffer with or without 2-mercaptoethanol prior to characterization by SDS-PAGE. Visualization by Coomassie staining showed non-reduced KS γ2 antibody to migrate as several species with molecular weights of approximately 150 kD. In contrast, the KSγ2h antibody migrated as a major band with an apparent molecular weight of 150 kD. When the KS γ2 antibody and KS γ2h antibody were reduced with mercaptoethanol before SDS-PAGE, an identical pattern of bands coπesponding to the heavy and light chains was observed for both the KSγ2 antibody and KSγ2h antibody.

[0147] Without wishing to be bound by theory, these observations suggest that the distinctly migrating species seen with KSγ2 were due to variations in disulfide bonding patterns. [0148] Stable clones of mammalian cells expressing KSγ2 antibody and KSγ2h antibody were also identified. Cell transfections were carried out under essentially identical conditions and similar numbers of stably transfected cells were cloned and tested for production of huFc KSγ2 antibody and KSγ2h antibody. The four best- expressing clones for KSγ2h antibody made about 114, 98, 85, and 49 micrograms of antibody per ml of tissue culture supernatant, while under the same conditions the four best-expressing clones for huFc KSγ2 antibody made about 36, 34, 15, and 13 micrograms of antibody per ml of tissue culture supernatant.

Example 7: Construction of a fusion protein including a complete antibody that contains IgG2 CHI. CH2, and CH3 domains and an IgGl hinge. [0149] In this Example, the usefulness of an antibody fusion protein in which the antibody moiety had a hybrid isotype was examined. [0150] The expression vector for huKSγ2-IL2, pdHL7-huKSγ2-IL2, was constructed by replacing the Smal-Xhol restriction fragment in pdHL7-huKSγ2 antibody, which contains the sequence GGGTAAATGA (SEQ ID NO: 24) followed by the Xhol sticky end, by the Smal-Xhol restriction fragment isolated from pdHL7- huKS-IL2. This latter fragment contains the sequence GGGTAAA followed immediately by the DNA sequence encoding the mature IL2 including the translation stop codon. The resultant expression vector, pdHL7-huKSγ2-IL2, was used to transfect cells for protein expression.

[0151] Similarly, the expression vector for huKSγ2h-IL2, pdHL7-huKSγ2h-IL2, was constructed by replacing the Smal-Xhol restriction fragment in pdHL7-huKSγ2h antibody, which contains the sequence GGGTAAATGA (SEQ ID NO: 24) followed by the Xhol sticky end, by the Smal-Xhol restriction fragment isolated from pdHL7- huKS-IL2 as described in the preceding paragraph. The resultant expression vector, pdΗL7-huKSγ2h-IL2, was used to transfect cells for protein expression.

Example 8: Characterization of the non-reduced state of a γ2-antibody fusion protein and a coπesponding γ2h antibody fusion protein.

[0152] For transient transfection, the plasmids encoding KS γ2-IL2 and KS γ2h- IL2 were introduced into mammalian cells by lipofection using Lipofectamine Plus (Life Technologies, Gaithersburg, MD) according to supplier's protocol.

[0153] In order to obtain stably transfected clones, plasmid DNAs were introduced into mouse myeloma NS/0 cells by electroporation as described in Example 2.

[0154] KS-IL2 fusion proteins with the γ2 and γ2h isotypes were characterized by SDS-PAGE as described in Example 6. Antibody fusion proteins in the conditioned media (10% serum) were captured on Protein A Sepharose (Repligen, Cambridge, MA) and then eluted by boiling in the protein sample buffer with or without 2-mercaptoethanol prior to characterization by SDS-PAGE. Visualization by Coomassie staining showed non-reduced KSγ2-IL2 to migrate as several species with molecular weights of approximately 180 kD. In contrast, the KSγ2h-IL2 migrated as a major band with an apparent molecular weight of about 180 kD. When the KSγ2- IL2 and KSγ2h-IL2 were reduced with mercaptoethanol before SDS-PAGE, an identical pattern of bands coπesponding to the heavy chain-IL2 and light chains was observed for both the KSγ2-IL2 and KSγ2h-IL2.

[0155] Without wishing to be bound by theory, these observations suggest that the distinctly migrating species seen with KSγ2-IL2 were due to variations in disulfide bonding patterns.

[0156] The modified γ2h immunocytokines are tested in animals and have extended half-lives with better efficacy in animal tumor models relative to original IgGγl -based immunocytokines. [0157] Stable clones of mammalian cells expressing KSγ2-IL2 and KSγ2h-IL2 were also identified. Cell transfections were carried out under essentially identical conditions and similar numbers of stably transfected cells were cloned and tested for production of KSγ2-IL2 and KSγ2h-IL2. The four best-expressing clones for KSγ2h- IL2 made about 52, 37, 31, and 30 micrograms of fusion protein per ml of tissue culture supernatant, while under the same conditions the four best-expressing clones for KSγ2-IL2 made about 31, 27, 27, and 17 micrograms of fusion protein per ml of tissue culture supernatant.

Example 9; Construction of plasmids expressing antibody fusion proteins with a hybrid isotype and a secondary mutation affecting fusion protein activity.

HuKSγ2-Ala-IL2 vs huKSγ2h-Ala-IL2

[0158] The expression vectors pdHL7-huKSγ2- Ala-IL2 and pdHL7-huKSγ2h-

Ala-IL2 (described above) were constructed by replacing the Smal-Xhol restriction fragment in pdHL7-huKSγ2 antibody and pdHL7-huKSγ2h antibody, respectively, which contains the sequence GGGTAAATGA (SEQ ID NO: 24) followed by the Xhol sticky end, by the coπesponding Smal-Xhol restriction fragment isolated from pdHL7-huKSγl-Ala-IL2. This latter fragment contains the sequence GGGTGCA followed immediately by the DNA sequence encoding the mature IL2 including the translation stop codon. The GCA encodes the lysine to alanine substitution at the junction of the fusion protein. The resultant vectors were used to transfect cells for the production of huKSγ2-Ala-IL2 and huKSγ2h-Ala-IL2. Example 10: Characterization of the non-reduced state of a γ2-antibodv fusion protein and a coπesponding γ2h antibody fusion protein caπying a secondary mutation affecting fusion protein function. [0159] For transient transfection, the plasmids encoding KSγ2-Ala-IL2 and KSγ γ2h-Ala-IL2 were introduced into mammalian cells by lipofection using Lipofectamine Plus (Life Technologies, Gaithersburg, MD) according to supplier's protocol.

[0160] Stable clones of mammalian cells expressing KSγ2-Ala-IL2 and KSγ2h- Ala-IL2 were also identified as in Example 2. Cell transfections were carried out under essentially identical conditions and similar numbers of stably transfected cells were cloned and tested for production of KSγ2-Ala-IL2 and KSγ2h-Ala-IL2. The three best-expressing clones for KSγ2h-Ala-IL2 made about 39, 38, and 29 micrograms of fusion protein per ml of tissue culture supernatant, while under the same conditions the three best-expressing clones for KSγ2-Ala-IL2 made about 22, 17, and 6 micrograms of fusion protein per ml of tissue culture supernatant. KS-Ala- IL2 fusion proteins with the γ2 and γ2h isotypes were characterized by SDS-PAGE as described in Example 6. KS-Ala-IL2 fusion proteins in the conditioned media (10% serum) were captured on Protein A Sepharose (Repligen, Cambridge, MA) and then eluted by boiling in the protein sample buffer with or without 2-mercaptoethanol prior to characterization by SDS-PAGE. Visualization by Coomassie staining showed non- reduced KSγ2-Ala-IL2 to migrate as several species. In contrast, the KSγ2h-Ala-IL2 migrated as one major species. When the KSγ2-Ala-IL2 and KSγ2h-Ala-IL2 are reduced with mercaptoethanol before SDS-PAGE, an identical pattern of bands coπesponding to the heavy chain-IL2 and light chains was observed for both the KSγ2-Ala-IL2 and KSγ2h-Ala-IL2.

[0161] Without wishing to be bound by theory, these observations suggest that the distinctly migrating species seen with KSγ2-Ala-IL2 were due to variations in disulfide bonding patterns. Example 11 : Expression of an antibody fusion protein with constant regions derived from different isotypes.

[0162] In some cases, it is desirable to construct an lg fusion protein in which different constant regions, in addition to the hinge region, are derived from different heavy chain isotypes. To examine the properties of this class of molecules, the following experiments were conducted.

[0163] The huKS(γl :CH1-H)( γ2:CH2-CH3)-Ala-IL2 protein, which is an antibody-IL2 fusion protein with the IgG heavy chain comprising the KS VH, the CHI and hinge regions from IgGl and CH2-CH3 regions from IgG2 (with the lysine to alanine substitution at the fusion junction, as described above) followed by IL-2, was expressed as follows. The expression vector pdHL7-huKS(γl :CH1-H)( γ 2:CH2- CH3)-Ala-IL2 was constructed by replacing the Hindlll-Afllll restriction fragment in pdHL7-KS γ2-Ala-IL2 with the coπesponding Hindlll-Afllll restriction fragment containing the CHI and hinge regions from pdHL7-KS-_L2. This expression vector was transfected into cultured mammalian cells as described above to yield an antibody fusion protein with a hybrid isotype.

[0164] The HuKS(γl :CHl-H)(γ2:CH2-CH3)(Ala)-IL2 hybrid and HuKSγ2(Ala)- IL2 antibody-cytokine fusion proteins were characterized by non-reducing SDS- PAGE as described in Example 6. The HuKS(γl:CHl-H)(γ2:CH2-CH3)(Ala)-_L2 hybrid isotype fusion protein migrated as a major band with a molecular weight of about 180kD. In contrast, the HuKSγ2(Ala)-IL2 antibody-cytokine fusion protein migrated as multiple bands in the 180 kD size range.

[0165] When the HuKS(γl :CHl-H)(γ2:CH2-CH3)(Ala)-IL2 hybrid and HuKS(γ2)(Ala)-IL2 antibody-cytokine fusion proteins were characterized by reducing SDS-PAGE, both proteins gave an identical pattern of bands, coπesponding to the light chain and the heavy chain-IL2 fusion polypeptides.

[0166] Without wishing to be bound by theory, it appeared that the HuKS(γ2)(Ala)-IL2 antibody-cytokine fusion protein existed in at least two different isomeric configurations, in which the difference was due to a difference in disulfide bonding pattern. Example 12: Expression of hybrid isotype antibody fusion proteins with a non- protein antigen specificity and a multi-subunit fusion partner. [0167] The 14.18 antibody binds to the glycolipid antigen GD2. Interleukin-12 is a heterodimeric cytokine that includes a p35 and p40 subunit, covalently attached by a disulfide bond.

[0168] As described above, use of a hybrid isotype, such as an IgGl/IgG2 hybrid, leads to enhanced assembly as compared to a natural isotype, such as IgG2. Enhanced assembly may be evidenced by increased expression levels.

[0169] In one case, 14.18(γ2)-Ala-IL12 and 14.18(γ2h)-Ala-IL12 expression plasmids were constructed and transiently transfected into cells in parallel under identical conditions as described above and in Gillies et al. (WO09929732). Human IL-12 was used. The level of protein in the tissue culture supernatant was about 40% higher from the cultures transfected with the 14.18(γ2h)-Ala-IL12 expression plasmid than from the cultures transfected with the 14.18(γ2)-Ala-IL12 expression plasmid. [0170] The 14.18(γ2)-Ala-IL12 and 14.18(γ2h)-Ala-IL12 expression plasmids were also stably transfected into cells as described above, and the four highest- expressing clones from each transfection were studied further. For 14.18(γ2h)-Ala- IL12, the average of the four best-expressing clones was about 45% higher than for the four best-expressing clones derived from the 14.18(γ2)-Ala-_L12 expression plasmid.

[0171] In a second case, 14.18(γ2)-IL12 and 14.18(γ2h)-IL12 expression plasmids were constructed and transiently transfected into cells in parallel under identical conditions as described above. Mouse IL-12 was used. The level of protein in the tissue culture supernatant was about 40% higher from the cultures transfected with the 14.18(γ2h)-ILl 2 expression plasmid than from the cultures transfected with the 14.18(γ2)-IL12 expression plasmid.

[0172] The 14.18(γ2)- IL12 and 14.18(γ2h)- IL12 expression plasmids using murine IL-12 were also stably transfected into cells as described above, and the four highest-expressing clones from each transfection were studied further. For 14.18(γ2h)-IL12, the average of the four best-expressing clones was about 35% higher than for the four best-expressing clones derived from the 14.18(γ2)-IL12 expression plasmid.

[0173] These results indicated that use of a hybrid isotype in a fusion protein led to enhanced expression, even when using a different antibody and different non-Ig moiety than in the previous examples. In this Example, the use of IL-12 as the non-Ig moiety is expected to sigmficantly complicate the assembly of the lg fusion protein, because IL-12 is a heterodimer. Nonetheless, the use of a hybrid isotype had a beneficial effect.

Example 13: Expression of hybrid isotype antibody fusion proteins using a hinge region from IgA. [0174] To construct an lg fusion protein with enhanced protease resistance, an IgA/IgG fusion protein is produced.

[0175] For example, an Fc-X fusion protein is constructed that contains the hinge region of human IgAl and the CH2 and CH3 regions of IgG2.

[0176] An example of the construction of an expression vector to produces an Fc- X fusion protein comprising the hinge region from human IgAl and the CH2 and CH3 regions of IgG2 is as follows. The plasmid pdCs-huFcγ2-Leptin from Example 1 is used as a vector.

[0177] The Aflll-Stul fragment containing the γ2 hinge exon in pdCs-huFcγ2- Leptin is replaced by the coπesponding Aflll-Stul fragment (SEQ ID NO: 25) containing the IgAl hinge exon, which is shown below:

(Aflll ) P S T P P T P S P S T P

CTTAAG T CCC TCA ACT CCA CCT ACC CCA TCT CCC TCA ACT CCA

P T P S P S C C H (SEQ ID NO : 26 ) (Stul )

CCT ACC CCA TCT CCC TCA TGC TGC CAC Ggtaagccagcccagrgcct [0178] Specifically, the following oligonucleo tides are synthesized:

Top strand: 5 ' -

TTAAGTCCCTCAACTCCACCTACCCCATCTCCCTCAACTCCACCTACCCCATCTCCC TCATGCTGCCACGGTAAGCCAGCCCAGG-3' (SEQ ID NO: 27)

Bottom strand: 5 ' -

CCTGGGCTGGCTTACCGTGGCAGCATGAGGGAGATGGGGTAGGTGGAGTTGAGGGAGA TGGGGTAGGTGGAGTTGAGGGAC-3' (SEQ ID NO: 28)

[0179] These oligonucleotides are annealed, and used to replace the coπesponding fragment in pdCs-huFcγ2-Leptin.

[0180] Since the Stul sites in the γ2 exon is C-methylated and the Stul restriction endonuclease is methylation sensitive, the plasmid is isolated from a DCM negative strain of bacteria before digestion with the Stul enzyme. The resultant pdCs-huFcγ2- Leptin with the hinge region from IgAl is designated pdCs-huFcαl/γ2-Leptin.

[0181] The plasmid pdCs-huFcαl/γ2-Leptin is transfected into eukaryotic cells arid an Fc-X protein of the form αl(hinge)-γ2(CH2, CH3)-leptin is expressed as a secreted protein. For example, the cells and methods of Example 2 are used. The resulting αl-γ2-leptin protein is purified, characterized, and found to possess leptin activity and to be relatively insensitive to protease cleavage in the hinge region.

Example 14: Expression of hybrid isotype antibody fusion proteins using components of IgG and a polyvalent immunoglobulin. [0182] To construct antibody fusion proteins with the effector functions of an IgG, such as IgGl or IgG3, and an increased valency, a hybrid isotype lg fusion protein using the CHI, hinge, and CH2 regions of an IgG and the CH3 and CH4 regions of an IgA or IgM are constructed. Figure 4 shows the structure of an IgG- IgM hybrid isotype fusion antibody. It is convenient to fuse a non-Ig moiety to the C- terminus of the IgA or IgM CH4 domain.

[0183] It is also convenient to truncate the IgA or IgM heavy chain before the most C-terminal cysteine, or to mutate this cysteine, especially when the IgG/IgA or IgG/IgM hybrid isotype fusion protein is expressed in the absence of a J chain. The normal function of this C-terminal cysteine is to form a disulfide bond with the J chain. It is often desirable to express an IgG/IgA or IgG/IgM hybrid isotype fusion protein in the absence of a co-expressed J chain, especially when fusing the non-Ig moiety to the C-terminus of the heavy chain.

[0184] For example, an IgG-IgM hybrid isotype fusion protein is constructed as follows. The plasmid pdHL7-huKSγl-IL2 is capable of expressing an antibody with variable regions that bind to Epithelial Cell Adhesion Molecule and constant regions from human IgGl . This plasmid contains a unique Ngo M IV site in the intron between the IgGl CH2 and CH3 coding sequences, and also contains a unique Xmal site at the junction between the coding sequences for the IgGl CH3 and the IL-2 moieties. A DNA fragment encoding the human IgM CH3 and CH4 sequences is generated by PCR amplification from human genomic DNA using the following primers:

5'-GCA GCCGGC CCTGAGCCTTGGCTTCCCAGAGCG-3' (SEQ ID NO: 29); and

5'- GCT CCCGGG TC AGTAGC AGGTGCCAGCTGTGTCG-3 ' (SEQ ID NO: 30)

[0185] Alternatively, the following primers are used; these have the feature that the C-terminal-most cysteine in the IgM moiety is mutated to a serine.

5'-GCA GCCGGC CCTGAGCCTTGGCTTCCCAGAGCG-3'; (SEQ ID NO: 31) and

5'- GCT CCCGGG TC AGTAGCTGGTGCCAGCTGTGTCG-3 '(SEQ ID NO: 32);

[0186] The resulting DNA fragment is cleaved with Ngo M IV and Xmal and is then directly or indirectly ligated to pdHL7-huKSγl-IL2 that has been cut with Ngo M IV and Xmal. For example, the PCR fragment encoding the IgM CH3 and CH4 domains is first subcloned into a vector such as the TA cloning vector (Invitrogen, Carlsbad CA), the sequence of the insert is verified, and then the insert is removed using Ngo M IV and Xmal and ligated into pdHL7-huKSγl-IL2. The amino acid sequence of a resulting hybrid isotype heavy chain, fused to human IL-2 is shown in SEQ ID 33. The IgM portion of the fusion protein is underlined. Because of naturally occuring polymoφhisms in human IgM heavy chains, it is also possible to generate closely related sequences that are functionally similar.

ASTKGPSVFPLAPSSKSTSGGTAALGCLVKDYFPEPVTVS WNSGALTSGVHTFPAVLQSSGLYSLSSWTVPSSSLGTQT YICNVNHKPSNTKVDKKVEPKSCDKTHTCPPCPAPELLGG PSVF FPPKPKDT MISRTPEVTCVWDVSHEDPEVKFN YVDGVEVHNAKTKPREEQYNSTYRWSVLTVLHQD LNGK EYKCKVSNKALPAPIEKTISKAKDQDTAIRVFAIPPSFAS IFLTKSTKLTCLVTDLTTYDSVTIS TRQNGEAVKTHTNI SESHPNATFSAVGEASICEDD NSGERFTCTVTHTDLPSP LKQTISRPKGVALHRPDVYLLPPAREQLN RESATITCLV TGFSPADVFVQWMQRGQPLSPEKYVTSAP PEPQAPGRYF AHSILTVSEEEWNTGETYTCWAHEALPNRVTERTVDKST GKPTLYNVSLVMSDTAGTCY

(SEQ ID NO: 33)

[0187] Many alternative DNA construction strategies are possible. For example, a DNA sequence encoding the IgM CH3 and CH4 domains may be generated by RT- PCR from an IgM-expressing cell line or cell population. The 3' PCR primer may be identical to that shown above, but the 5' primer incoφorates an artificial 5' splice site between the Ngo M IN site and the beginning of the IgM CH3 coding sequence.

[0188] Eukaryotic cells are then transformed with the resulting plasmid, pdHL7- huKSγl/μ-IL2, and the resulting protein is expressed and secreted from the cells. For example, the cells and methods described in the previous Examples are used. The purified protein is tested, for example by electron microscopy or by size-exclusion chromatography, and found to have a multimeric structure. The purified protein is further studied, for example by peptide mapping, and it is found that the cysteine in the sequence ...ASICEDD.... (SEQ ID NO: 34) forms an interchain disulfide bond with other IgG/IgM hybrid isotype subunits. An example of such a linkage is illustrated in Figure 4. In addition to the pentameric structure of Figure 4, the final form of the IgG/IgM hybrid isotype protein may be hexameric or some other type of multimeric structure.

[0189] The resulting IgG/IgM hybrid isotype fusion protein is found to bind to Fcγ receptors. For example, the fusion protein binds to J774 cells in a manner that is competitive with human IgG.

Example 15: Expression of hybrid isotype antibody using components of IgGl and IgG4. [0190] The puφose of the experiments in Examples 15-17 was, in part, to demonstrate the utility of hybrid isotype antibodies in improving the assembly of molecules consisting primarily of IgG4. In general, a commonly prefeπed form of an IgG molecule is the H₂L₂ form, with two heavy chains and two light chains. However, a significant fraction of IgG4 antibodies are synthesized as HL "half molecules" with one heavy chain and one light chain. Angal et al. (1993; Molec.

Immunol. 30:105) described a serine to proline substitution that decreased the amount of "half molecules" found in secreted human IgG4.

[0191] This example compares the assembly properties of three antibodies containing the same V regions: an IgG4 form, a form with a hinge from IgGl and the CHI, CH2, and CH3 domains from IgG4, and an IgG4 form with a mutation in the hinge region (Angal et al., ibid.).

[0192] Using standard plasmid construction techniques, a plasmid termed pdHL7- KS-γ4 was constructed. This plasmid is identical to the plasmid pdHL7-huKSγ2 described in Example 5, except that it encodes an IgG4 heavy chain constant region instead of an IgG2 heavy chain constant region.

[0193] To construct pdHL7- KS-γ4h, the pdHL7-KS-γ2h plasmid was grown in a dcm(-) strain of E. coli, plasmid DNA isolated, and the 60 bp Pstl-Stul fragment encoding the modified IgGl hinge region was isolated and used to replace the coπesponding fragment in pdHL7- KS-γ4. [0194] For comparison, the IgG4 form with a mutation in the hinge region (Angal et al., ibid.) was constructed as follows. An oligonucleotide duplex encoding the γ4 hinge with the serine to proline substitution, and a 5' Pstl sticky end and a 3' Stul blunt end, was designed as follows:

Pstl E S K Y G P P C P P C P (SEQ ID NO: 35) GAG TCC AAA TAT GGT CCC CCA TGC CCA CCT TGC CCA GGT AAG

ACGT CTC AGG TTT ATA CCA GGGGGT ACG GGT GGA ACG GGT CCATTC Stul CCAACCCAGG (complement of SEQ ID NO: 36) GGTTGGGrCC (SEQ ID NO: 36)

[0195] This oligonucleotide duplex was used to replace the coπesponding DNA fragment in ρdHL7-KS-γ4 to give pdHL7-KS- γ4(S to P).

[0196] The pdHL7- KS-γ4, pdHL7- KS-γ4h, and pdHL7- KS-γ4(S to P) plasmids were transfected into mammalian cells according to standard procedures, for example, as described in Lo et al., (1998; Protein Engineering 11 :495-500). From the supernatants of transfected cells, the antibody proteins encoded by pdHL7- KS-γ4, pdHL7- KS-γ4h, and pdHL7- KS-γ4(S to P) were purified and characterized by SDS gel electrophoresis in which reduced and non-reduced samples were compared.

[0197] By examining non-reduced molecules, it was found that the KS-γ4 antibody population existed as about 50% HL "half-molecules" and 50% H₂L₂ molecules. In contrast, the KS-γ4h antibody population and the KS-γ4 (S to P) antibody population existed almost entirely as H₂L₂ molecules. The proportion of HL half-molecules was about the same in the KS-γ4h antibody population and the KS-γ4 (S to P) antibody population. When reduced molecules were examined, the pattern of heavy and light chains seen with KS-γ4, KS-γ4h, and KS-γ4(S to P) were indistinguishable.

Example 16: Expression of hybrid isotype antibody fusion proteins using components of IgGl and IgG4.

[0198] The plasmid ρdHL7- KS-γ4-IL2 was described in Gillies et al. (Cancer Research [1999] 59:2159). This plasmid encodes an antibody fusion protein with V regions that recognize the EpCAM antigen and contain the heavy chain of IgG4 with interleukin-2 fused at its C-terminus.

[0199] By recombinant DNA procedures analogous to those used in Example 15, the plasmids pdHL7- KS-γ4h-IL2 and pdHL7- KS-γ4 (S to P)-_L2 were constructed. The plasmids pdHL7- KS-γ4-IL2, pdHL7- KS-γ4h-IL2 and pdHL7- KS-γ4 (S to P)- IL2 transfected into mammalian cells, and the coπesponding proteins were expressed and purified according to standard procedures, for example, as described in Lo et al. (ref). From the supernatants of transfected cells, the fusion proteins encoded by pdHL7- KS-γ4-IL-2, pdHL7- KS-γ4h-IL-2, and pdHL7- KS-γ4(S to P)-IL-2 were purified and characterized by SDS gel electrophoresis in which reduced and non- reduced samples were compared. [0200] By examining non-reduced molecules, it was found that the KS-γ4-lL-2 fusion protein population existed as about 50% HL "half-molecules" and 50% H₂L₂ molecules. In contrast, the KS-γ4h-IL-2 fusion protein population and the KS-γ4 (S to P)-E -2 fusion protein population existed almost entirely as H₂L₂ molecules. The proportion of HL half-molecules was about the same in the KS-γ4h-IL-2 fusion protein population and the KS-γ4(S to P)-IL-2 fusion protein population. When reduced molecules were examined, the pattern of heavy and light chains seen with KS-γ4-IL-2, KS-γ4h-IL-2, and KS-γ4(S to P)-IL-2 were indistinguishable.

Example 17: Expression of hybrid isotype Fc fusion proteins using components of IgGl and IgG4. [0201] To create an set of plasmids for expression of fusion proteins containing a hinge region, a CH2 and CH3 domain, and a non-Ig moiety, in which the lg moieties were derived from IgG4 and IgGl, the following steps were undertaken.

[0202] First, using procedures analogous to those described in Example 1 , an expression vector encoding the Fc region from IgG4 was created by replacing the IgGl-derived DNA sequences of pdCs-huFcγl (Lo et al., [1998] Protein Engineering 11 :495-500) with sequences encoding the coπesponding part of IgG4. Specifically, the following oligonucleotide

E S K Y G P P C P S C (SEQ ID NO: 37)

C TTA AGC GAG TCC AAA TAT GGT CCC CCA TGC CCA TCA TGC (SEQ ID NO: 38)

which encodes an Aflll site at its 5' end, was used as a 5' primer to amplify a DNA segment encoding the IgG4 hinge, CH2, and CH3 regions. The 3' primer contained a Xhol site at its 5' end, analogously to the primer used in Example 1 to amplify the IgG2 Fc region. The resulting Aflll-Xhol fragment was inserted into XhoI+Aflll-cut pdCs-huFcγl to create pdCs-huFcγ4.

[0203] To facilitate the insertion of nucleic acids encoding non-IgG moieties at the C-terminus of Fcγ4, a Smal site was created by introducing the following Leu to Pro change near the C-terminus of the Fcγ4 coding region in pdCs-huFcγ4, as follows:

S L G K STOP S P G K STOP

(SEQ ID NO: 39) (SEQ ID NO: 40)

TCT CTG GGT AAA TGA was changed to TCC CCG GGT AAA TGA. (SEQ ID NO: 41) (SEQ ID NO: 42)

[0204] Standard site-directed mutagenesis techniques were used to introduce the Smal site into pdCs-huFcγ4.

[0205] To create a plasmid, termed pdCs-huFcγ4h, encoding the modified IgGl hinge followed by the IgG4 CH2 and CH3 domains, the Stul-Xhol fragment of pdCs- huFcγ2h (Example 1), which encodes the IgG2 CH2 and CH3 domains, was replaced with the coπesponding fragment from pdCs-huFcγ4. In this construction, each parental plasmid was derived from a dcm(-) strain of E. coli. The pdCs-huFcγ4 plasmid was completely digested with Xhol and partially digested with Stul, since there is an additional Stul site in the γ4-encoding DNA, to yield fragments of about 300, 500 and 800 base pairs. The approximately 800 base pair fragment was used to replace the coπesponding fragment in pdCs-huFcγ2h.

[0206] An IFNβ cDNA was cloned by PCR using sense primer CCCGGGT ATG AGC TAC AAC TTG CTT GGA TTC (SEQ ID NO: 43), where ATG is the N- terminal residue of the mature protein and CCCGGG is an introduced Smal restriction site, and reverse primer CTCGAG TCA GTT TCG GAG GTA ACC TGT AAG (SEQ ID NO: 44), where TCA is the anticodon of the translation stop codon and CTCGAG is an introduced Xhol restriction site. The template DNA was pLGl 17R (Taniguchi et al., [1980] Proc. Nat. Acad. Sci. USA 77:5230-5233) and was purchased from the American Type Culture Collection (ATCC number 31903). After sequence verification, the cloned Smal-Xhol fragment was ligated to the Smal and Xhol sites of the pdCs-huFcγ4 expression vector to give pdCs- huFcγ4-IFNβ. An analogous pdCs- huFcγ4h-IFNβ expression plasmid was also constructed by similar methods.

[0207] To construct the pdCs- huFcγ4(S to P)-IFNβ expression plasmid, an oligonucleotide duplex encoding the γ4 hinge with the serine to proline substitution, and a 5' Aflll sticky end and a 3' Stul blunt end was synthesized as follows:

Aflll E S K Y G P P C P P C P (SEQ ID NO:

45)

TTAAGC GAG TCC AAATAT GGT CCC CCATGC CCA CCT TGC CCA GGT G CTC AGG TTT ATACCA GGG GGT ACG GGT GGA ACG GGT CCA

Stul AAGCCAACCCAGG (SEQ ID NO: 46) TTC GGTTGGGTCC (complement ofSEQ IDNO: 46)

[0208] This DNA was used to replace the coπesponding DNA fragment in pdCs- Fc-g4-IFNβ to give pdCs-huFcγ4(S to P)-IFNβ.

[0209] The plasmids pdCs-huFcγ4-IFNβ, pdCs-huFcγ4h-IFNβ, and pdCs- huFcγ4(S to P)-_FNβ were each transfected into mammalian cells. The expressed Fc- containing proteins were purified and examined by reducing and non-reducing SDS gel electrophoresis. It was found that a significant proportion of the huFcγ4-IFNβ protein expressed from mammalian cells was monomeric half-molecules rather than dimeric, normal antibodies. However, the expressed huFcγ4h-IFNβ, and huFcγ4(S to P)-IFNβ fusion proteins were almost entirely coπectly assembled dimers. For both the huFcγ4h-IFNβ, and huFcγ4(S to P)-IFNβ fusion proteins, the proportion of monomer was about the same.

[0210] Taken together, the results of Examples 15, 16, and 17 indicate that hybrid isotype antibodies and lg fusion proteins consisting primarily of IgG4 but having the modified IgGl hinge have superior assembly properties as compared to coπesponding proteins derived entirely from IgG4. In addition, the results of Examples 15-17, in combination with other Examples, illustrate that the improved assembly of hybrid isotype proteins can be manifested in several different ways. For example, in some cases a hybrid isotype antibody or fusion protein shows reduced aggregation, as compared to a coπesponding single-isotype antibody or lg fusion protein; and in other cases the hybrid isotype antibody or fusion protein shows enhanced coπect oligomerization, as compared to a coπesponding single-isotype antibody or lg fusion protein.

Example 18: Expression of hybrid isotype Fc fusion proteins using components of IgGl and IeG4.

[0211] To generate an Fc-erythropoietin fusion protein that is minimally aggregated when expressed from mammalian cells, the following expression plasmid was constructed using standard molecular biology techniques. An Xmal-Xhol DNA fragment containing a form of the human erythropoietin coding sequence with mutations resulting in the amino acid substitutions His32Gly, Cys33Pro, Tφ88Cys, and Pro90Ala, as disclosed in WO01/36489, was used. The coπesponding protein sequence is shown in SEQ ID NO: 47: APPRLICDSRVLERY LEAKEAENITTGCAEGPS NENITVPDTKvTJFYAWKRMEVGQQAVEVWQGLAL LSEAVLRGQALLVNSSQPCEGLQLHVDKAVSGLRSLTTLLRALGAQKEAISPPDAASAAPLRTITADTF RKLFRVYSNFLRGK KLYTGEACRTGDR

[0212] This Xmal-Xhol DNA fragment was inserted into a plasmid vector that encodes a hinge region from IgGl and a CH2 and CH3 region from IgG2 that was essentially identical to the vector constructed in Example 1, except that there were two sets of mutations that resulted in amino acid substitutions in the region of the CH3 C-terminus, such that the sequence at the junction of the CH3 C-terminus and the Epo N-terminus is as follows: . . . . TQKSATATPGA-APPRLI . . . .(SEQ ID NO: 48)

[0213] The first set of mutations, which change the sequence KSLSLSPG (SEQ ID NO: 49) of the IgG2 CH3 region to KSATATPG (SEQ ID NO: 45), is disclosed in U.S. Patent Application Serial No. 60/280,625. The effect of the substitution of Leu- Ser-Leu-Ser (position 3 to position 6 of SEQ ID NO: 49) with Ala-Thr-Ala-Thr (position 3 to position 6 of SEQ ID NO: 50) is to remove potential human non-self T- cell epitopes that may arise because the junction between human Fc and human erythropoietin contains non-self peptide sequences. The second set consisting of the single amino acid substitution K to A at the C-terminal amino acid of the CH3 region, is disclosed in U.S. Patent Application Serial No. 09/780,668. [0214] The resulting plasmid was transfected into NS/0 cells and the Fc-Epo fusion protein was expressed and purified according to the procedures of Examples 1 and 2. After purification based on binding to protein A, the huFcγ2h-huEpo protein containing the IgG2 CH3 and erythropoietin substitutions described above was characterized by size exclusion chromatography and found to consist of 97% monomer and 90% monomer in two independent preparations. The huFcγ2h-huEpo protein containing the IgG2 CH3 and erythropoietin substitutions described above was found to be about as active, on a molar basis, as human erythropoietin in a cell- based assay that measured the ability of an erythropoietin protein to stimulate TF-1 cell division. The assay was performed as described in WO01/36489.

[0215] In addition, fusions of non-mutant human erythropoietin to the C-terminus of an Fc region consisting of either IgGl(hinge-CH2-CH3), IgG2(hinge-CH2-CH3), or IgGl(hinge)-IgG2(CH2-CH3) were characterized. Expression plasmids comprising non-mutant human Fc sequences and non-mutant erythropoietin sequences were constructed analogously to the plasmid described above and in

Example 1. NS/0 cells were transfected with the Fcγl-Epo, Fcγ2-Epo, and Fcγ2h-Epo expression plasmids, and stable clones were isolated after screening an approximately equal number of clones for each plasmid. The best-producing clones yielded 50 μg/ml for Fcγl-Epo, 20 μg/ml for Fcγ2-Epo, and 120 μg/ml for Fcγ2h-Epo. Equivalents

[0216] The invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. The foregoing embodiments are therefore to be considered in all respects illustrative rather than limiting on the invention described herein. Scope of the invention is thus indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are intended to be embraced therein.

Incorporation by Reference [0217] All patents, patent applications, and scientific publications mentioned herein above are incoφorated by reference into this application in their entirety.

Claims

What is claimed is: 1. A fusion protein comprising an immunoglobulin moiety fused to a non- immunoglobulin moiety, wherein said immunoglobulin moiety comprises a first domain from a first antibody isotype and a second domain from a second antibody isotype.

2. A fusion protein of Claim 1 , wherein said immunoglobulin region comprises a hinge region from a first antibody isotype and a constant region from a second antibody isotype.

3. The fusion protein of Claim 2, wherein said constant region is a CH2 domain.

4. The fusion protein of Claim 3, wherein said CH2 domain is from a human IgG2 heavy chain, and said hinge region contains three or fewer cysteine residues.

5. The fusion protein of Claim 3, wherein said CH2 domain is from a human IgG4 heavy chain.

6. The fusion protein of Claim 2, wherein the hinge region is a non-IgG2 hinge region.

7. The fusion protein of Claim 2, wherein the hinge region is derived from human IgGl.

8. The fusion protein of Claim 3-6, wherein the hinge region has two or fewer cysteines.

9. The fusion protein of Claim 1, wherein said immunoglobulin domain lacks an antigen-binding site.

10. The fusion protein of Claim 1, wherein said immunoglobulin domain comprises an antigen-binding site.

11. The fusion protein of Claim 1 , wherein the non-Ig moiety is fused by genetic engineering to the C-terminus of a heavy chain of the lg domain.

12. The fusion protein of Claim 1, wherein the non-Ig moiety is fused by genetic engineering to the N-terminus of a heavy chain of the lg domain.

13. The fusion protein of Claim 1 , wherein the non-Ig moiety is fused to a light chain polypeptide.

14. The fusion protein of Claim 2, wherein the hinge region is from a human IgA.

15. The fusion protein of Claim 1, wherein said first immunoglobulin domain is a CH2 domain and said second immunoglobulin domain is a CH3 domain.

16. The fusion protein of Claim 15, wherein the first antibody isotype is an IgG and the second antibody isotype is an IgM or an IgA.

17. An antibody or lg fusion protein comprising an IgG2 antibody moiety in which the IgG2 hinge has been mutated two contain two cysteines.

18. An immunoglobulin moiety comprising a V domain directed against a tumor- associated antigen, said immunoglobulin moiety comprising a first domain from a first antibody isotype and a second domain from a second antibody isotype.

19. The protein of claim 18, comprising an inactive Fc receptor I binding site and a hinge region with three or fewer cysteines.

20. A nucleic acid encoding a fusion protein of Claim 1, 17, orl8.

21. A stably transfected cell expressing a protein of Claim 1 , 17, or 18.

22. A method for improving the assembly of an antibody or lg fusion protein, the method comprising the steps of:

a. replacing the hinge region of said protein with a different hinge region; and, b. testing the assembly of the altered protein

23. The method of Claim 22, wherein said protein and said altered protein are produced by genetic engineering techniques.

24.The fusion protein of claim 1, wherein the immunoglobulin moeity comprises an Fc heavy chain region and the non-immunoglobulin moeity comprises an erythropoietin molecule.

25. A nucleic acid encoding the fusion protein of claim 24.

26. The fusion protein of claim 1, wherein the non-immunoglobulin moeity comprises an erythropoietin molecule.

27. A method for improving the production of an antibody or lg fusion protein from a cell, the method comprising the steps of:

a. replacing the hinge region of said protein with a different hinge region; and

b. testing the production of the altered protein.

28. The method of Claim 27, wherein said protein and said altered protein are produced by genetic engineering techniques.